Skip to content
Back to search

cve-intel

Local security intelligence for CVE triage - 34 tools, 25 sources, risk scores and an MCP server for AI agents, all in native Rust. Private by default.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 11 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/Fazmin/cve-intel with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/Fazmin/cve-intel'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/fazmin--cve-intel/badge.svg)](/listings/fazmin--cve-intel)

HTML

<a href="/listings/fazmin--cve-intel"><img src="/api/listings/fazmin--cve-intel/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Download a release

Download the file for your OS from the [latest release](https://github.com/Fazmin/cve-intel/releases/latest). The desktop app, CLI, MCP server and REST API are all **one executable**, with nothing else to install.

| OS | File | How to run | |---|---|---| | Windows | `CVE-Intel_<version>_windows-x64-setup.exe` (or `…-portable.exe`) | Run the installer, then open **CVE Intel** | | macOS (Apple Silicon and Intel) | `CVE-Intel_<version>_macos-universal.dmg` | Drag **CVE Intel** to Applications | | Linux | `CVE-Intel_<version>_linux-x64.AppImage` (or `.deb`) | `chmod +x` the file and run it |

Each file has a `.sha256` checksum next to it. Builds are not code-signed yet, so Windows SmartScreen and macOS Gatekeeper warn on first launch. On macOS, right-click the app and choose **Open**.

Build from source

# Desktop app (debug)
cargo build -p cve-intel-desktop
./target/debug/cve-intel-desktop

# Installable bundle (.dmg / .exe / .AppImage)
cargo install tauri-cli --version "^2" --locked
cd apps/desktop/src-tauri && cargo tauri build

# Headless binary only (servers, Docker)
cargo build --release -p cve-server

Docker

Runs the headless MCP server over HTTP, plus the REST API:

docker build -t cve-intel .
docker run --rm -p 127.0.0.1:8000:8000 --env-file .env cve-intel

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 1,443 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. <h1 align="center">CVE Intel</h1>
  2. <p align="center">
  3. Know which vulnerabilities matter, and what to do about them.<br/>
  4. A private desktop app for vulnerability triage and scanning, written in native Rust, with a CLI, REST API and MCP server for developers and AI agents,<br/>
  5. and a fleet mode that watches every server you run from one place.<br/>
  6. 42 MCP tools, 46 data sources, SBOMs in and out, on-device AI.
  7. </p>
  8. <p align="center">
  9. <a href="https://www.rust-lang.org"><img src="https://img.shields.io/badge/Rust-native-orange?style=flat-square&logo=rust" alt="Rust"/></a>
  10. <a href="https://tauri.app"><img src="https://img.shields.io/badge/Tauri-2-blue?style=flat-square" alt="Tauri 2"/></a>
  11. <a href="https://modelcontextprotocol.io"><img src="https://img.shields.io/badge/MCP-Compatible-blue?style=flat-square" alt="MCP compatible"/></a>
  12. <a href="#license"><img src="https://img.shields.io/badge/License-MIT%20OR%20Apache--2.0-green?style=flat-square" alt="License: MIT OR Apache-2.0"/></a>
  13. </p>
  14. <p align="center">
  15. <a href="#what-is-cve-intel">What it is</a> ·
  16. <a href="#getting-started">Getting started</a> ·
  17. <a href="#features">Features</a> ·
  18. <a href="#fleet-mode">Fleet mode</a> ·
  19. <a href="#install">Install</a> ·
  20. <a href="#local-ai">Local AI</a> ·
  21. <a href="#data-sources">Data sources</a> ·
  22. <a href="#developer-notes">Developer notes</a> ·
  23. <a href="#citing-and-attribution">Citing</a>
  24. </p>
  25. <p align="center">
  26. <a href="assets/remediation-lab-demo.mp4"><img src="assets/remediation-lab-demo.webp" width="900" alt="Sped-up screen recording of CVE Intel on macOS: triage of CVE-2024-6387 and its How to fix it card, then the Remediation Lab: the plan for openssh on Ubuntu 22.04 with Ubuntu's own build first, ticking the fix and opening its runbook, ticking Ubuntu's mitigation instead and watching the SSVC decision move from Attend to Track, saving that as a decision and finding it in the ledger, then PAN-OS 10.2.8 with its hotfix and workaround, RHEL 8 as not affected, log4j-core, and Plan fix from a scan of lodash 4.17.20"/></a>
  27. </p>
  28. <p align="center"><sub>Sped-up demo on macOS (<a href="assets/remediation-lab-demo.mp4">video</a>, or the <a href="assets/remediation-lab-fixed.mp4">full-length recording</a>): from a triage result to a remediation plan for what you run, its runbook, a mitigation tried in What if and saved as a decision, and Plan fix from a scan. A second demo, of the landscape, decisions, scans and the watchlist, is under <a href="#features">Features</a>.</sub></p>
  29. ## What is CVE Intel?
  30. Every known software vulnerability gets a public ID, such as CVE-2024-3400.
  31. For each one, security, IT and development teams face the same question:
  32. *does this affect us, and do we fix it now or can it wait?* Answering it
  33. usually means checking NVD, CISA, FIRST, GitHub and vendor advisories by
  34. hand, one browser tab at a time.
  35. ### Why it matters now
  36. CVE publication keeps rising
  37. ([48,185 in 2025](https://jerrygamblin.com/2026/01/01/2025-cve-data-review/),
  38. 20.6% more than 2024, and
  39. [over 86,000 by 27 October 2026](https://craci.com/resources/CVE-publication-trends)),
  40. and in April 2026
  41. [NIST said](https://www.nist.gov/news-events/news/2026/04/nist-updates-nvd-operations-address-record-cve-growth)
  42. the National Vulnerability Database cannot keep up, so it will enrich CVEs in
  43. CISA's Known Exploited Vulnerabilities (KEV) catalog, federal-use software
  44. and critical software first, and move its unenriched backlog from before
  45. 1 March 2026 to "Not Scheduled". Meanwhile,
  46. [Mandiant's M-Trends 2026](https://cloud.google.com/blog/topics/threat-intelligence/m-trends-2026)
  47. estimates the mean time to exploit at −7 days, so exploitation routinely
  48. begins before a patch is released, and exploits were the most common initial
  49. infection vector for the sixth year running (32% of intrusions).
  50. ### What it does
  51. Give CVE Intel a CVE ID, a project folder, a lockfile or SBOM, a container
  52. image, the software on this computer, your own domains, or a suspicious IP
  53. address, file hash or domain. It queries up to 46 public sources, including
  54. CISA KEV, FIRST's EPSS and ENISA's EU Vulnerability Database alongside NVD,
  55. and returns what the vulnerability is, a risk score from 0 to 100, whether
  56. attackers are already using it, the evidence behind each signal and what to
  57. do next. It also shows what changed this week across those sources, and
  58. watches the vendors, products and packages you run.
  59. - **46 data sources** for vulnerability, exploit, network and threat
  60. intelligence, listed under [Data sources](#data-sources). Many need no
  61. API key.
  62. - **42 MCP tools** for AI agents and coding assistants. The same program is
  63. also a command line and a REST API (see
  64. [Developer notes](#developer-notes)).
  65. - **SBOMs in and out.** It reads CycloneDX and SPDX SBOMs along with
  66. lockfiles, and **Save SBOM** (or `scan deps . --format cyclonedx`) writes
  67. any scan as a CycloneDX 1.5 SBOM with the findings, their risk signals and
  68. your recorded decisions.
  69. - **Local.** There is no CVE Intel account or hosted service. Lookups go
  70. straight from your computer to the public sources, and the AI assistant
  71. runs on a local model unless you choose a remote provider. Fleet mode runs
  72. on a server you own.
  73. It is for security analysts, IT and operations teams, developers checking
  74. their dependencies, and anyone learning vulnerability management.
  75. ### One machine or a whole fleet
  76. The desktop app checks the computer it runs on and whatever you give it. To
  77. watch your servers as well, run **fleet mode**: a small agent on each machine
  78. sends what is installed there to a CVE Intel server you run. That server
  79. checks every machine when its software changes and again every day, records
  80. what became new, riskier or fixed, and answers *which of our machines have
  81. this CVE?* The desktop app's **Fleet** view shows it all from your own
  82. computer.
  83. <p align="center">
  84. <picture>
  85. <source media="(prefers-color-scheme: dark)" srcset="assets/fleet-mode-dark.svg"/>
  86. <img src="assets/fleet-mode-light.svg" width="900" alt="How fleet mode works: an agent on each machine (web01 on Ubuntu 22.04, app01 on Debian 12, db01 on Ubuntu 24.04) reads its packages, apps and lockfiles offline and sends the inventory over HTTPS to your central server, cve-intel fleet serve. Only the server looks the software up in the public sources (OSV, NVD, CISA KEV, EPSS and exploit feeds). It checks each inventory when it changes and every day, records what is new, riskier or fixed, and answers which machines a CVE is on. The desktop app's Fleet view, webhook alerts to Slack, Mattermost or anything else, and the REST API and command line read the results with a viewer token. Agents never listen, and the machines need no internet access, only a route to your server."/>
  87. </picture>
  88. </p>
  89. [Fleet mode](#fleet-mode) under Features has screenshots and a demo, and
  90. [the fleet guide](docs/fleet.md) shows how to set it up.
  91. ### What it helps you do
  92. - **Decide whether to patch now:** a 0–100 risk score, a plain
  93. recommendation and the evidence behind it.
  94. - **Fix it on what you run:** the **Remediation Lab** ranks the fixes and
  95. workarounds for your distribution, product or package, each with its
  96. source.
  97. - **Put a list of CVEs in order:** paste up to 10 from a scanner report and
  98. get the patch order.
  99. - **Check your dependencies:** drop a project folder, lockfile or SBOM and
  100. see the vulnerable packages, riskiest first, with the version that fixes
  101. each.
  102. - **Check what you run and ship:** this computer's apps and OS, a container
  103. image, or what public sources know about your own domains.
  104. - **See what changed this week** across CISA KEV, EPSS, ENISA, the exploit
  105. databases and CISA advisories.
  106. - **Watch the products you run** and hear about new CVEs, KEV listings,
  107. exploits and fixes.
  108. - **Remember what you decided** and hear when the reasons stop holding.
  109. - **Watch every server you run:** which machines a new CVE is on, what
  110. changed on each since yesterday, and which ones stopped reporting.
  111. - **Search by product or KEV:** every CVE for a product and version, or
  112. CISA's exploited list filtered by vendor, ransomware use or date.
  113. - **Investigate something suspicious:** an IP address, file hash, domain or
  114. URL.
  115. - **Ask in plain English:** the assistant looks up the facts first and
  116. shows its sources.
  117. - **Give AI coding assistants the same knowledge** through MCP.
  118. ### Terms you'll see
  119. | Term | What it means |
  120. |---|---|
  121. | **CVE** | A public ID for one known vulnerability, such as CVE-2021-44228 (Log4Shell). |
  122. | **CVSS** | A 0–10 severity score: how bad the flaw is if someone exploits it. It says nothing about how likely an attack is. |
  123. | **EPSS** | The probability that a vulnerability will be exploited in the next 30 days, published daily by FIRST. |
  124. | **CISA KEV** | The US government's catalog of vulnerabilities known to be exploited by attackers. A CVE on this list is urgent. |
  125. | **ENISA EUVD** | The EU vulnerability database. Its own "exploited" flag is independent of CISA KEV and can raise the SSVC Exploitation input. |
  126. | **Exploit / PoC** | Public code that demonstrates or carries out an attack. A Metasploit module means the attack is packaged and easy to use. |
  127. | **SSVC** | A decision method from CISA and CERT/CC that ends in one of four actions: *Act*, *Attend*, *Track\** or *Track*. |
  128. | **SBOM / lockfile** | A list of every package your software uses, with exact versions. |
  129. | **MCP** | Model Context Protocol, the standard way AI assistants call outside tools. |
  130. The **risk score** combines these signals: severity (CVSS), the chance of
  131. exploitation (EPSS), confirmed exploitation (CISA KEV) and public exploit
  132. code. Anything on CISA KEV is always CRITICAL. See [Risk score](#risk-score)
  133. for the formula.
  134. > CVE Intel helps you decide what to patch. It does not probe, port-scan or
  135. > attack anything, and it does not apply fixes. Scans read what is already
  136. > on your machine (dependency files, the software installed on this
  137. > computer, a container image you choose) and send only package and product
  138. > names and versions to OSV.dev and NVD. The exposure check is passive: it
  139. > asks public sources (Shodan, certificate transparency logs) what they
  140. > already know about domains and addresses you own, and sends nothing to
  141. > them. Results are only as current as the public sources.
  142. ## Getting started
  143. You don't need an account or any API keys to start. Most people are
  144. looking at their first result within a couple of minutes:
  145. 1. [Download the app](#install) for your operating system and open it. The
  146. **Landscape** view loads first: what changed this week across the public
  147. sources.
  148. 2. Open **CVE Triage**, type a CVE ID such as `CVE-2024-3400` and press
  149. **Triage**, or click one of the examples under the search box. Press
  150. **Watch CVE** (or **Watch** next to the vendor) to be told when it
  151. changes.
  152. 3. Drag a project folder onto the window to check its dependencies, or open
  153. **Scan** and choose **This computer** to check the software installed on
  154. this machine. Findings come back ranked by real-world risk, with the
  155. version that fixes each one.
  156. 4. Press `⌘K` (`Ctrl+K` on Windows and Linux) to open **Quick Find**. Paste
  157. anything there (a CVE, IP address, file hash, domain, URL or
  158. `package:ecosystem:version`) and CVE Intel picks the right lookup.
  159. 5. *Optional:* in **Settings → Data source API keys**, add free API keys for
  160. faster lookups and more exploit data. See [API keys](#api-keys).
  161. 6. *Optional:* open **Local Model** and download a model to use **Ask AI**
  162. with answers that never leave your computer.
  163. 7. *Optional:* open **MCP Server** to connect your AI coding assistant.
  164. The buttons at the bottom of the sidebar switch between the light and dark
  165. themes and make the text larger or smaller. The version you are running is
  166. shown below them.
  167. Everything up to [Developer notes](#developer-notes) is about the desktop
  168. app. To use CVE Intel from a terminal, a script, a CI pipeline or another
  169. program, go straight to that section.
  170. ## Features
  171. <p align="center">
  172. <a href="assets/core-functions-demo.mp4"><img src="assets/core-functions-demo.webp" width="900" alt="Sped-up screen recording of CVE Intel on macOS: the threat landscape, CVE Triage of CVE-2024-3400 and recording a decision, the decision ledger with stale, expired and review-due examples, a dependency scan of a demo project and its attack chains, and the watchlist's recorded changes"/></a>
  173. </p>
  174. <p align="center"><sub>Sped-up demo on macOS (<a href="assets/core-functions-demo.mp4">video</a>): the threat landscape, triage and recording a decision, the decision ledger, a scan and its attack chains, and the watchlist.</sub></p>
  175. The short version of each feature is below. For every option and caveat,
  176. see [Features in detail](docs/features.md).
  177. ### Threat landscape
  178. What changed recently across independent sources: new CISA KEV entries,
  179. CVEs ENISA marks exploited that KEV doesn't list, the biggest EPSS rises,
  180. new Metasploit modules and PoCs, new critical CVEs, Project Zero's
  181. in-the-wild zero-days and CISA advisories. Your watchlist comes first, then
  182. CVEs that several panels agree on. **Copy as Markdown** for a weekly note.
  183. ![Threat landscape for the last 30 days: tiles for 10 unread watchlist changes, 35 CVEs added to CISA KEV, one CVE ENISA reports exploited that KEV does not list, new exploits and PoCs, EPSS risers, new critical CVEs, this year's zero-days and CISA advisories, above the watchlist panel with a chip for each watched item and its latest changes, such as two Ivanti CVEs added to CISA KEV on 30 October and EPSS rises for Fortinet, Ivanti and Citrix CVEs](assets/screenshots/landscape.png)
  184. ### Watchlist
  185. Watch the vendors, products, packages and CVEs you run. Each refresh
  186. records what changed: new CVEs, KEV listings, EPSS jumps, new exploits,
  187. CVSS changes and newly available fixes. While the app is open it refreshes
  188. every 6 hours (change it in **Settings**). The list stays on this machine.
  189. ![What the watchlist recorded for its Fortinet, Palo Alto Networks, Ivanti and Citrix items in October: six CVEs added to CISA KEV between 4 and 30 October, and four EPSS jumps, one from 4.0% to 91.5%, each with its date and a Mark read button](assets/screenshots/watchlist.png)
  190. ### Decisions
  191. Record what you decided on a CVE (patch now, defer until a date, accept,
  192. not affected, mitigated) and the assumptions behind it, such as "not in
  193. CISA KEV" or "not internet-facing". Every refresh re-checks them:
  194. - **Stale:** an assumption broke. The CVE entered KEV, an exploit or fix
  195. appeared, or EPSS crossed your threshold.
  196. - **Expired / review due:** a deferral passed its date, or a manual
  197. assumption passed its review date.
  198. - **Calibration:** how your past calls held up, such as how many CVEs you
  199. held back later entered KEV or gained an exploit, and how soon.
  200. Export the ledger as OpenVEX or CycloneDX VEX. It stays on this machine;
  201. AI agents can read it but never change it.
  202. ### CVE Triage
  203. One lookup across NVD, CVE.org, EPSS, CISA KEV, ENISA EUVD and the exploit
  204. databases gives a 0–100 risk score, a recommendation and the evidence
  205. behind it. Deep triage adds CISA's SSVC decision. Two panels go further:
  206. - **Who says what:** the sources side by side, and where they disagree
  207. (scores 2+ points apart, NVD hasn't analysed it, a distribution says
  208. "not affected"). Marked **Sources agree**, **Sources disagree** or **Thin
  209. evidence**.
  210. - **Exploit speed:** how soon an exploit is likely (**Hours**, **Days**,
  211. **Weeks**), estimated from the public fix commit, the weakness type and
  212. the CVSS vector, with every reason shown.
  213. Neither changes the risk score.
  214. ![Deep CVE Triage for CVE-2024-3400 in the dark theme: a 100/100 CRITICAL score, CISA KEV and ENISA exploited flags, a weaponized exploit, the exploit-speed tile and CISA's SSVC "Act" decision](assets/screenshots/triage-dark.png)
  215. ### Ask in plain language
  216. Ask *"Which is worse: CVE-2024-3400 or CVE-2023-44487?"* CVE Intel runs the
  217. lookups first, then the model explains the results and names the tools it
  218. used. Small local models can get details wrong, so check anything you will
  219. act on in **CVE Triage**.
  220. ### Compare and prioritise
  221. Paste 2 to 10 CVE IDs and get a patch order: CISA KEV first, then the risk
  222. score, with ties broken by ransomware use, EPSS and exploit maturity. Each
  223. CVE says why it sits above the next. The side-by-side table highlights the
  224. riskier values, and a summary shows what the CVEs share (same product,
  225. reachable without sign-in, public exploits).
  226. ![Compare CVEs running the KEV vs. CVSS example: Log4Shell and HTTP/2 Rapid Reset, both in CISA KEV, ranked above curl's CVSS 9.8 SOCKS5 flaw, with the reason each one sits above the next](assets/screenshots/compare.png)
  227. ### Remediation plans
  228. Say what you run (a distribution package, a vendor product and version, or
  229. a library) and the **Remediation Lab** merges every source's advice into
  230. one ranked plan:
  231. - **Fixes:** your distribution's own build first (Ubuntu and Red Hat
  232. backport fixes), then the vendor's update or the upstream version.
  233. - **Workarounds and mitigations** in the vendor's words, with general
  234. controls ranked last.
  235. - **Every option names its source.** Nothing is made up, and a source that
  236. could not be reached is listed as a gap, not as "no advisory".
  237. Open it with **Open in Remediation Lab** on a triage result, or **Plan
  238. fix** on a scan finding.
  239. ```bash
  240. cve-intel remediate CVE-2024-6387 --distro ubuntu:jammy --package openssh@1:8.9p1-3ubuntu0.6
  241. cve-intel remediate CVE-2024-3400 --product "PAN-OS@10.2.8"
  242. cve-intel remediate CVE-2020-28500 --package npm:lodash@4.17.20
  243. ```
  244. ![Remediation Lab for CVE-2024-6387 with openssh 1:8.9p1-3ubuntu0.6 on Ubuntu jammy: a fix exists; Ubuntu's own build 1:8.9p1-3ubuntu0.10 is first and marked "For what you run", citing Ubuntu Security, OSV.dev and USN-6859-1, with the packages to install and why it sits above the next option; Ubuntu's mitigation is ticked second; beside them the What if panel shows SSVC moving from Attend to Track with exposure controlled and mission impact low, what remains exposed, and the decision it amounts to, "Defer until 2026-12-15"](assets/screenshots/remediation-lab.png)
  245. **What if.** Tick the options you are considering to see what stays exposed
  246. and how the SSVC decision changes, then **Save as decision…** (`--apply
  247. <option id> --record` on the command line).
  248. **Runbooks.** **Runbook…** (`--runbook`) turns the options you chose into
  249. steps: what to run, how to check it took, and how to back it out. CVE
  250. Intel never runs them.
  251. | Written by | What leaves your computer |
  252. |---|---|
  253. | Fixed rules (the default) | Nothing |
  254. | The local model (adds a summary) | Nothing |
  255. | A cloud provider under **Remote AI** (`--runbook --ai`) | The CVE, the kind of system, the options you chose and their public advisory text |
  256. A model's steps are checked against the plan's evidence, and risky commands
  257. (piping a download into a shell, turning off a firewall) are rejected.
  258. ![Runbook dialog for that fix, made by fixed rules with no AI: the apt-get commands that upgrade openssh-client and openssh-server, the version each package should then be at, a reminder to restart what uses the package, and the dpkg-query command that checks it took, with buttons to copy the runbook or save it as Markdown](assets/screenshots/remediation-runbook.png)
  259. ### Scan dependencies
  260. Drop a project folder on the window, or paste a lockfile, SBOM or package
  261. list. CVE Intel finds the npm, Yarn, Python, Rust, Go, PHP, Ruby and Maven
  262. files and CycloneDX or SPDX SBOMs, and checks the packages with OSV.dev.
  263. Only package names and versions leave your computer.
  264. - **Ranked by real risk:** Log4Shell in an old `log4j-core` comes first; a
  265. CVSS 9.8 nobody exploits does not. Each finding shows the version that
  266. fixes it.
  267. - **History:** scan again to see what is new, resolved or riskier. Tick
  268. **Monitor** to rescan daily, or **Watch all affected** to add the
  269. packages to the watchlist.
  270. - **Chains:** links findings into attack chains and names the **break
  271. point**, the one fix that cuts the most chains.
  272. - **Save SBOM:** writes the scan as a CycloneDX 1.5 SBOM with the findings,
  273. their risk signals and your decisions.
  274. To run the same scan in a CI pipeline, see
  275. [CI and automation](#ci-and-automation).
  276. ![Scan of a demo project with four dependency files: 16 of 16 packages affected, 4 findings in CISA KEV, ranked by risk with Log4Shell in log4j-core first (CISA KEV, ransomware and weaponized-exploit signals), then HTTP/2 Rapid Reset in golang.org/x/net and a second Log4j flaw, each with the version that fixes it and Watch, Plan fix and Decide buttons](assets/screenshots/scan.png)
  277. ![Scan History after the demo project was scanned again: log4j-core upgraded to 2.17.1 and lodash to 4.17.21 and ws added, so the row shows 3 new and 6 resolved findings, with the Monitor switch, Scan again and Forget, and each resolved CVE marked as fixed by an upgrade](assets/screenshots/scan-history.png)
  278. ### Scan this computer
  279. **Scan → This computer** checks the OS and installed apps: from the
  280. registry on Windows, Applications and Homebrew on macOS, and the package
  281. database on Linux. Apps CVE Intel recognises are checked with NVD; Linux
  282. packages with OSV.dev (Arch with the Arch Linux Security Tracker, matched
  283. locally). Apps it doesn't recognise are listed as not checked. **List
  284. installed software** shows what would be checked without contacting
  285. anything.
  286. ### Scan a container image
  287. **Scan → Container image** checks a `docker save` or OCI archive without
  288. Docker. Its OS packages and any lockfiles inside are checked with OSV.dev.
  289. Distributions OSV doesn't cover are listed as not checked, never reported
  290. clean.
  291. ```bash
  292. docker save myapp:latest -o myapp.tar
  293. cve-intel scan image myapp.tar --fail-on kev
  294. ```
  295. ### Check your public exposure
  296. **Scan → Public exposure** shows what public sources already know about
  297. domains and addresses you own. It finds your host names in certificate
  298. transparency logs, then asks Shodan what services, versions and CVEs have
  299. been seen on them. It is passive: CVE Intel never connects to or
  300. port-scans the targets.
  301. ```bash
  302. cve-intel scan exposure example.com --authorized
  303. ```
  304. ### Databases and data sources
  305. Every upstream source, what it provides, whether it needs a key and how
  306. long results are cached. **Check connectivity** probes them all live.
  307. ![Databases view: 46 upstream sources, 30 of them free without a key, the local cache and downloaded model, and a connectivity check with 35 of 46 sources reachable and 7 waiting on an API key](assets/screenshots/databases.png)
  308. ### MCP server for AI agents
  309. Give Claude Code, Cursor, VS Code, Claude Desktop, Windsurf or any MCP
  310. client CVE Intel's 42 tools. Start the server from the app and copy the
  311. setup steps, already filled in. Agents connect over HTTP (local only, port
  312. 8000 by default) or launch their own copy over stdio, and use your saved
  313. API keys. To set it up by hand, see [MCP server setup](#mcp-server-setup).
  314. ![MCP Server view in the dark theme with the server running on port 8000, setup steps for Claude Code over HTTP, and the 42 tools it offers](assets/screenshots/mcp.png)
  315. ### Fleet mode
  316. <p align="center">
  317. <a href="assets/fleet-demo.mp4"><img src="assets/fleet-demo.webp" width="900" alt="Screen recording of the Fleet view on macOS: entering the central server's address and a viewer token and connecting; the overview of 5 hosts with 751 findings, 10 in CISA KEV; searching which hosts have CVE-2021-44228, which finds app01 with log4j-core 2.14.1; opening app01 and ticking In KEV only, which leaves Log4Shell, its incomplete fix and Spring4Shell in the billing project; then Plan fix, which opens the Remediation Lab with the upgrade to log4j-core 2.15.0 for that version"/></a>
  318. </p>
  319. <p align="center"><sub>Demo on macOS, 38 seconds (<a href="assets/fleet-demo.mp4">video</a>): connecting the Fleet view to a central server, asking which machines have Log4Shell, opening one and planning the fix for the version it runs.</sub></p>
  320. One CVE Intel server watching many machines. A small agent
  321. (`cve-intel-agent`, 2 to 4 MB) on each server reports what is installed
  322. there: its OS packages and apps, and the dependency files of the projects
  323. you name. Your central server checks every machine's inventory against the
  324. sources when it changes and again every day, even when nothing on the
  325. machine changed, and records what became new, riskier or fixed. Ask it
  326. which machines a CVE is on, hear when one stops reporting, and send alerts
  327. to a webhook. The agents need no internet access, never listen for
  328. connections, and hold no vulnerability data.
  329. The server and the agents run from the command line. The desktop app's
  330. **Fleet** view shows the fleet from your own computer: every machine and
  331. its state, each one's findings, which machines a CVE is on and what
  332. changed, with **Plan fix** opening the Remediation Lab for the version that
  333. machine runs. See [the fleet guide](docs/fleet.md) to set it up.
  334. ![Fleet view connected to a central server: 5 hosts, all reporting; 751 findings, 11 critical and 9 high; 10 in CISA KEV across 4 hosts; no problems. The hosts table lists app01 and ci01 on Debian 12, web01 on Ubuntu 22.04, and db01 and edge01 on Ubuntu 24.04, each with its group, state, findings by risk, how many are in KEV, and when it last reported and was checked](assets/screenshots/fleet.png)
  335. ![One host in the Fleet view: app01 on Debian 12, group prod, cve-intel-agent 0.3.0 reporting every 15 minutes, 247 findings with 5 in KEV. Filters for risk, KEV only, and new or riskier findings, then its system target's findings, riskiest first: CVE-2023-4911 in glibc 2.36-9+deb12u1, in CISA KEV and weaponized, fixed in 2.36-9+deb12u3, each finding with Plan fix and Hosts buttons](assets/screenshots/fleet-host.png)
  336. ### Also included
  337. - **Quick Find** (`⌘K` or `Ctrl+K`): paste a CVE, IP, hash, domain, URL or
  338. `package:ecosystem:version` and the right lookup runs.
  339. - **Fixed versions:** the version that fixes every known vulnerability in a
  340. package. Quick Find `lodash:npm:4.17.15` to see it.
  341. - **Product and KEV search:** every CVE for a product and version, and
  342. filters over the CISA KEV catalog.
  343. - **MITRE ATT&CK mapping** for KEV-listed CVEs, plus inferred techniques,
  344. labelled as inferred.
  345. - **IP, hash and domain intelligence** from AbuseIPDB, GreyNoise, Shodan,
  346. VirusTotal, MalwareBazaar, crt.sh and more.
  347. - **Activity:** a searchable local history of your lookups and of the tool
  348. calls agents made.
  349. - **Local Model:** install the runtime and download a model in one click.
  350. ## Common uses
  351. | Situation | What to do in the app |
  352. |---|---|
  353. | What happened this week? | Open **Landscape**, or ask *"What's new this week?"* in **Ask AI**. |
  354. | A new CVE is in the news. Do we drop everything? | Enter it in **CVE Triage**, or ask *"Should we patch CVE-2024-3400 this weekend?"* in **Ask AI**. |
  355. | We run Fortinet (or another vendor). Tell me when something new hits it. | Add the vendor on **Watchlist**, or press **Watch** next to the vendor on a triage result. |
  356. | How do we fix this CVE on Ubuntu 22.04, or on the PAN-OS version we run? | Press **Open in Remediation Lab** on the triage result and say what you run. The plan puts the fix for your release or branch first and names its sources. |
  357. | The fix can't go in this week, or there isn't one yet. What do we do meanwhile? | In **Remediation Lab**, tick the vendor's workaround to see what it leaves exposed, then **Save as decision…**. A deferral is re-checked until its date, and if no fix existed you hear when one ships. **Runbook…** gives you the steps. |
  358. | We decided not to patch something yet. Tell me if that stops being safe. | Press **Record decision** on the triage result and tick the reasons. **Decisions** marks it stale the moment one of them breaks. |
  359. | A scanner reported 40 findings. Which come first? | Paste the IDs into **Compare CVEs**, up to 10 at a time. |
  360. | Checking dependencies before a release. | Drop the project folder on **Scan**, or paste the lockfile or SBOM. Press **Watch all affected** to follow them, or **Monitor** in History to have it scanned again daily. |
  361. | Is anything installed on this laptop being exploited? | **Scan → This computer**, then **Check**. Findings in CISA KEV come first. |
  362. | Is the image we ship affected by a new KEV entry? | `docker save` it and drop the `.tar` on **Scan → Container image**. |
  363. | What does the internet see on our servers? | **Scan → Public exposure** with your domain; passive, nothing connects to your hosts. |
  364. | Log4Shell is back in the news. Which of our servers have it? | With [fleet mode](#fleet-mode) set up, enter the CVE under **Fleet → Which hosts have a vulnerability?**, then **Plan fix** on a host's finding. |
  365. | Which version of a library fixes everything? | Open **Quick Find** and paste `lodash:npm:4.17.15`. |
  366. | What is known about the Tomcat version we run? | Ask *"Which CVEs affect Apache Tomcat 9.0.80?"* in **Ask AI**.¹ |
  367. | Is anything from a vendor being exploited right now? | Ask *"Which Fortinet vulnerabilities are in CISA KEV?"* in **Ask AI**.¹ |
  368. | An unknown IP appears in firewall or login logs. | Paste it into **Quick Find**. |
  369. | EDR flags an unknown file. | Paste its SHA-256 hash into **Quick Find**. |
  370. | You work in an AI coding tool. | Open **MCP Server**, start the server and copy the setup steps for your agent. |
  371. | You are new to vulnerability management. | Ask *"Explain EPSS and CISA KEV like I'm new."* |
  372. ¹ For these questions the model has to pick the right lookup itself, which
  373. larger local models and cloud providers do more reliably. The same searches
  374. are always available from the terminal (see [Recipes](#recipes)).
  375. ## Install
  376. Download the file for your OS from the
  377. [latest release](https://github.com/Fazmin/cve-intel/releases/latest). The
  378. desktop app, CLI, MCP server and REST API are all **one executable**, with
  379. nothing else to install.
  380. | OS | File | How to run |
  381. |---|---|---|
  382. | Windows | `CVE-Intel_<version>_windows-x64-setup.exe` (or `…-portable.exe`) | Run the installer, then open **CVE Intel** |
  383. | macOS (Apple Silicon and Intel) | `CVE-Intel_<version>_macos-universal.dmg` | Drag **CVE Intel** to Applications |
  384. | Linux | `CVE-Intel_<version>_linux-x64.AppImage` (or `.deb`) | `chmod +x` the file and run it |
  385. Each file has a `.sha256` checksum next to it. Builds are not code-signed
  386. yet, so Windows SmartScreen and macOS Gatekeeper warn on first launch. On
  387. macOS, right-click the app and choose **Open**.
  388. To build it yourself, run the headless server in Docker, or run the desktop
  389. app in a container and use it from your browser (no install), see
  390. [Build from source and Docker](#build-from-source-and-docker).
  391. ## Local AI
  392. **Ask AI** uses a small language model that runs on your computer, so your
  393. questions stay private. Before the model runs, CVE Intel finds the CVE IDs,
  394. IPs, hashes and domains in your question, runs the matching lookups and
  395. gives the model the results as facts to rely on. If no model is running,
  396. Ask returns those lookup results directly, so it is useful before any
  397. download finishes.
  398. Open the **Local Model** view to install the llama.cpp runtime and download
  399. a model, one click each (files go under `~/.cve-intel/`). The default is
  400. Qwen 2.5 3B Instruct, a ~1.9 GB download that needs 6 GB of RAM; the 1B and
  401. 1.5B models need 4 GB and Qwen 2.5 7B needs 10 GB. On Apple silicon the view
  402. also offers two MLX models that run on the GPU and answer faster. The model
  403. server listens on `127.0.0.1` only. Already running Ollama, LM Studio or
  404. your own llama-server? See
  405. [Use your own model server](#use-your-own-model-server).
  406. ### Remote providers (opt-in)
  407. For stronger answers, open **Configure → Remote AI** and choose OpenAI,
  408. Anthropic, Google Gemini, or any **OpenAI-compatible** endpoint: a LiteLLM
  409. proxy, vLLM, Azure OpenAI, OpenRouter, Together, Groq, Ollama behind HTTPS…
  410. Paste a key (and for an OpenAI-compatible endpoint, its base URL), then pick
  411. a model. **Save & use** enables the provider and tests it; **Save without
  412. switching** keeps your current choice. You can switch back to the local
  413. model at any time.
  414. | Provider | Default model | Environment variables |
  415. |---|---|---|
  416. | OpenAI | `gpt-4o` | `OPENAI_API_KEY` |
  417. | Anthropic | `claude-opus-5` | `ANTHROPIC_API_KEY` |
  418. | Google Gemini | `gemini-2.5-flash` | `GEMINI_API_KEY` |
  419. | OpenAI-compatible endpoint (LiteLLM, vLLM, …) | (the server's model name) | `CUSTOM_LLM_BASE_URL`, `CUSTOM_LLM_API_KEY` (optional) |
  420. **Models** come from the provider itself: once a key is saved, the model
  421. list is fetched live from the provider's `/models` endpoint, so it shows
  422. every model your key can use (for OpenAI and Gemini, embedding, speech and
  423. image models are left out). Press **↻** to refresh it, or choose **Other**
  424. to type any model ID. Until a key is saved, a few suggestions are shown.
  425. The same settings work from a terminal:
  426. ```bash
  427. cve-intel ai list # every provider, where it sends requests, what it still needs
  428. cve-intel ai set anthropic --key # hidden prompt
  429. cve-intel ai models anthropic # every model the key can use, fetched live
  430. cve-intel ai set anthropic --model claude-opus-5 --use
  431. cve-intel ai set custom --url https://litellm.example.com/v1 --key --model claude-sonnet-5 --use # LiteLLM, vLLM, …
  432. cve-intel ai test # one tiny request to the provider in use
  433. cve-intel ai use local # back to the on-device model
  434. ```
  435. `--key` reads the key from a hidden prompt (or stdin when piped), never from
  436. the command line.
  437. - **Off by default.** When a remote provider is selected, each question and
  438. the lookup results gathered for it are sent to that endpoint under its
  439. operator's terms. The Ask view shows a banner while this is on. A
  440. [remediation runbook](#remediation-plans) is written by the provider only
  441. when you ask for one, after the dialog (or the terminal) has shown
  442. exactly what would be sent.
  443. - **Where requests go.** OpenAI, Anthropic and Gemini use built-in HTTPS
  444. addresses. An OpenAI-compatible endpoint must be `https://`; plain
  445. `http://` is accepted only to this machine (`localhost`, `127.0.0.1`). A
  446. saved key is tied to its URL: changing the URL forgets the key, so it is
  447. never sent to a new host without being entered again. A key saved while
  448. the URL comes from `CUSTOM_LLM_BASE_URL` is saved with that URL, and
  449. `CUSTOM_LLM_API_KEY` is only sent to `CUSTOM_LLM_BASE_URL`. Requests to a
  450. loopback address skip `HTTPS_PROXY`.
  451. - **Keys stay local.** Keys are stored in `~/.cve-intel/settings.db` with
  452. owner-only permissions. They are never shown again in the UI and never
  453. logged. You can use environment variables instead. See
  454. [SECURITY.md](SECURITY.md).
  455. - **No vendor SDK.** Every provider here offers an OpenAI-compatible chat API
  456. with tool calling. Anthropic describes its compatible endpoint as intended
  457. mainly for evaluation; for Claude-specific features, use CVE Intel through
  458. MCP in a Claude client.
  459. ## Data sources
  460. The **Databases** view lists all 46 sources, and the CLI, REST API and MCP
  461. server report the same list.
  462. | Category | Sources |
  463. |---|---|
  464. | Vulnerability intelligence | NVD, NVD change history, CVE.org, CISA Vulnrichment, EPSS (FIRST) and its daily score files, CISA KEV, ENISA EUVD, VulnCheck NVD++, VulnCheck KEV, MSRC, Red Hat, Ubuntu, and CSAF advisories from Red Hat (VEX), SUSE (VEX), Cisco, Siemens and CISA ICS |
  465. | Exploit intelligence | GitHub PoC search, PoC-in-GitHub (nomi-sec), Exploit-DB, Metasploit modules, Nuclei templates, GitHub commits and pull requests (fix size for the exploit-speed estimate), Project Zero 0day In the Wild |
  466. | Network intelligence | AbuseIPDB, GreyNoise, Shodan, crt.sh, CIRCL Passive DNS |
  467. | Threat intelligence | VirusTotal, MalwareBazaar, ThreatFox, URLScan, CISA advisories (RSS), CIRCL Vulnerability-Lookup sightings, Ransomwhere, CIRCL hashlookup |
  468. | DevSecOps and reference | OSV.dev, Arch Linux Security Tracker, GitHub Advisories, MITRE ATT&CK (STIX), CTID KEV → ATT&CK mappings, MITRE CAPEC, HIBP Pwned Passwords |
  469. Many tools need no API key at all: EPSS and its daily files, KEV, ENISA
  470. EUVD, CVE.org and CISA Vulnrichment, OSV, the Arch Linux Security Tracker, CVSS, Exploit-DB, Metasploit,
  471. PoC-in-GitHub, Project Zero, CISA advisories, CIRCL sightings, ATT&CK,
  472. CAPEC, crt.sh, Ransomwhere, hashlookup, Pwned Passwords, and NVD at a
  473. lower rate limit. Large feeds (EPSS daily files, Exploit-DB, Metasploit,
  474. ATT&CK, CAPEC, the CTID mappings) are downloaded once, reduced to a small
  475. index, and refreshed daily or every few hours.
  476. ### Risk score
  477. ```
  478. Risk = CVSS×0.20 + EPSS×0.35 + KEV×0.30 + PoC×0.15 (boosters up to ×1.27, capped at 100)
  479. ```
  480. - **KEV override:** a CVE in CISA KEV is confirmed exploited, so it is always
  481. **CRITICAL** with a score of at least 76. Only CISA KEV earns the KEV
  482. points and the override.
  483. - EPSS measures the probability of exploitation. CVSS is treated as severity,
  484. not likelihood.
  485. - **PoC** is the strongest public exploit evidence. A Metasploit exploit
  486. module counts as weaponized (the full 15 points); Exploit-DB entries and
  487. VulnCheck XDB exploits count as public exploits.
  488. - Deep triage also returns the **SSVC v2** CISA Deployer decision:
  489. `Act`, `Attend`, `Track*` or `Track`. Its inputs come from CISA's published
  490. SSVC assessment when there is one. Because that assessment is a snapshot,
  491. newer evidence can raise Exploitation: a CISA KEV, VulnCheck KEV or ENISA
  492. EUVD exploited listing, or an EPSS of 0.90 or more, makes it `active`.
  493. EUVD and VulnCheck do not change the risk score.
  494. - Bands: 0–25 LOW, 26–50 MEDIUM, 51–75 HIGH, 76–100 CRITICAL.
  495. - This is scoring version 1.1, reported as `scoring_version` in results.
  496. ## API keys
  497. Keys are optional. Each one unlocks more data or raises a rate limit. In the
  498. app, open **Settings → Data source API keys**, press **Get a key** to open
  499. the sign-up page, paste the key and press **Save**. The key is used on the
  500. next lookup, with no restart needed.
  501. | Key | What it unlocks | Get it |
  502. |---|---|---|
  503. | NVD API key | 10× faster CVE lookups | [nvd.nist.gov](https://nvd.nist.gov/developers/request-an-api-key) |
  504. | GitHub token (no scopes) | Exploit search, Nuclei templates and advisory fixed versions (5,000 instead of 60 requests/hour) | [github.com/settings/tokens](https://github.com/settings/tokens) |
  505. | VulnCheck token | VulnCheck KEV (exploitation reports and XDB exploit links), and backup CVE data if NVD is down | [vulncheck.com](https://www.vulncheck.com/community) |
  506. | AbuseIPDB key | IP abuse score | [abuseipdb.com](https://www.abuseipdb.com/api) |
  507. | GreyNoise key | Whether an IP is harmless background noise | [docs.greynoise.io](https://docs.greynoise.io/docs/using-the-greynoise-community-api) |
  508. | Shodan key | Open ports and services on an IP | [account.shodan.io](https://account.shodan.io) |
  509. | VirusTotal key | 70+ antivirus verdicts for a file hash | [virustotal.com](https://www.virustotal.com/gui/my-apikey) |
  510. | URLScan key | Higher limits for URL checks | [urlscan.io](https://urlscan.io/docs/api/) |
  511. | abuse.ch Auth-Key | MalwareBazaar and ThreatFox lookups | [auth.abuse.ch](https://auth.abuse.ch/) |
  512. | CIRCL passive DNS login | Historical DNS records (access by request) | [circl.lu](https://www.circl.lu/services/passive-dns/) |
  513. Keys are stored in `~/.cve-intel/settings.db` with owner-only permissions
  514. and shared by the app, CLI, MCP server and REST API on the same machine. A
  515. key saved in the app takes precedence over an environment variable. Free
  516. tiers and sign-up steps are set by each vendor, so check their pages. To
  517. manage keys from a terminal or with environment variables, see
  518. [Terminal configuration](#terminal-configuration).
  519. ## Privacy and security
  520. - All network traffic is outbound HTTPS: to the data sources, to the cloud
  521. AI provider if you opt in to one, and for the model and runtime
  522. downloads you start (Hugging Face, the llama.cpp release on GitHub and,
  523. for MLX, `uv` and a Python 3.12 build from GitHub plus `mlx-lm` from
  524. PyPI).
  525. - API keys are never logged, cached, shown in the UI or returned by any tool.
  526. - Private and reserved IPs are rejected before any network lookup.
  527. - The local model server and the app's MCP server bind to loopback only.
  528. - SQLite files are created on first use: `cache.db`, a disposable lookup
  529. cache; `~/.cve-intel/settings.db`, which holds keys, provider settings
  530. and the watchlist auto-refresh interval; and `~/.cve-intel/watch.db`,
  531. the local watchlist (items, snapshots and change events). Clearing the
  532. cache never touches settings or the watchlist.
  533. - While the app is open, an optional background task can refresh the
  534. watchlist (every 6 hours by default). It only contacts the public
  535. sources needed for the items you watch, and you can turn it off in
  536. **Settings**. Nothing runs while the app is closed.
  537. - While the **Fleet** view is connected to a central server, the app asks
  538. that server for the fleet's state when you open or refresh the view, and
  539. once a minute to count new changes. It sends the viewer token and nothing
  540. about your computer. The address and the token are stored in
  541. `settings.db`, and **Disconnect** removes them.
  542. - The desktop app keeps an activity log in `~/.cve-intel/activity.db`
  543. (owner-only). It records what you look up and ask, the answers, setting
  544. and model changes, the tools agents call through the app's MCP server, and
  545. which hosts each action contacted. API keys are never recorded, only key
  546. names. Upstream requests are stored as host and path, without query
  547. strings. Open **Activity** to search it, pause it or clear it. It keeps
  548. the latest 20,000 entries. The CLI and `cve-intel mcp` record nothing.
  549. See [SECURITY.md](SECURITY.md) for the full security model and how to report
  550. a vulnerability.
  551. ## Developer notes
  552. Everything in this section is for using CVE Intel from a terminal, scripts,
  553. CI pipelines, other programs or AI agents, and for building it yourself. The
  554. desktop app needs none of it.
  555. - [The `cve-intel` executable](#the-cve-intel-executable): where it is and how one file runs every mode
  556. - [Command line (CLI)](#command-line-cli): every command, with [recipes](#recipes) for common tasks
  557. - [CI and automation](#ci-and-automation): fail a build, upload SARIF to GitHub code scanning, JSON output
  558. - [REST API](#rest-api): endpoints and examples
  559. - [MCP server setup](#mcp-server-setup): connect AI agents by hand
  560. - [Fleet mode](docs/fleet.md): a central server and agents on other machines
  561. - [Terminal configuration](#terminal-configuration): keys, environment variables, your own model server
  562. - [Build from source and Docker](#build-from-source-and-docker)
  563. - [Architecture](#architecture) and [Development and tests](#development-and-tests)
  564. ### The `cve-intel` executable
  565. The desktop app and the command-line tool are the same file. Double-clicking
  566. it opens the app; running it with any argument starts the CLI, the MCP
  567. server or the REST server instead.
  568. | OS | Executable |
  569. |---|---|
  570. | macOS | `/Applications/CVE Intel.app/Contents/MacOS/cve-intel` |
  571. | Windows (default per-user install) | `%LOCALAPPDATA%\CVE Intel\cve-intel.exe` |
  572. | Linux | `/usr/bin/cve-intel` (`.deb`), or the downloaded `.AppImage` |
  573. ```bash
  574. # macOS
  575. "/Applications/CVE Intel.app/Contents/MacOS/cve-intel" triage CVE-2024-3400
  576. # Linux
  577. ./CVE-Intel_<version>_linux-x64.AppImage triage CVE-2024-3400
  578. # Windows (default per-user install)
  579. "%LOCALAPPDATA%\CVE Intel\cve-intel.exe" triage CVE-2024-3400
  580. ```
  581. > On Windows the executable is a windowed program, so the terminal does not
  582. > wait for it. To see the output in order, use
  583. > `start /wait "" "…\cve-intel.exe" triage CVE-2024-3400`, or in PowerShell
  584. > `& "…\cve-intel.exe" … | Out-Host`. MCP clients start it directly and are
  585. > not affected.
  586. For servers and containers there is also a headless build without the GUI
  587. (`cargo build --release -p cve-server`, which produces
  588. `target/release/cve-intel`). From a source checkout, replace `cve-intel`
  589. with `cargo run -p cve-server --`. The examples below assume `cve-intel` is
  590. on your `PATH`.
  591. ### Command line (CLI)
  592. ```bash
  593. cve-intel triage CVE-2024-3400 --depth deep
  594. cve-intel ask "Should we patch CVE-2021-44228?" --model qwen2.5-3b
  595. cve-intel quick-find 185.220.101.34
  596. cve-intel scan deps . # every lockfile, manifest and SBOM here, ranked by risk
  597. cve-intel scan deps . --fail-on kev # exit 3 if anything is in CISA KEV
  598. cve-intel fix lodash npm 4.17.15 # version that fixes every known vulnerability
  599. cve-intel remediate CVE-2024-6387 --distro ubuntu:jammy --package openssh@1:8.9p1-3ubuntu0.6
  600. cve-intel product confluence_data_center --vendor atlassian --version 8.5.0
  601. cve-intel kev-search --vendor microsoft --ransomware-only --epss
  602. cve-intel landscape --days 7 # what's new; --markdown for a ticket
  603. cve-intel watch add vendor fortinet
  604. cve-intel watch refresh && cve-intel watch changes --unread
  605. cve-intel decide record CVE-2025-12345 --verdict defer --until 2026-12-31 --auto --note "waiting for the vendor"
  606. cve-intel decide list --state needs_attention # decisions whose reasons stopped holding
  607. cve-intel decide calibration # how your decisions held up
  608. cve-intel compare CVE-2024-3400 CVE-2023-44487
  609. cve-intel password-check # hidden prompt; only a 5-character hash prefix is sent to HIBP
  610. cve-intel sources --probe # every data source, with live status
  611. cve-intel tls-check # HTTPS diagnosis: proxy, CA store, certificate chain, verdict
  612. ```
  613. | Command | What it does |
  614. |---|---|
  615. | `triage <CVE> [--depth quick\|standard\|deep]` | Risk score, recommendation, KEV, EPSS and exploit evidence; `deep` adds SSVC |
  616. | `lookup <CVE>` | The NVD and CVE.org record |
  617. | `search <keyword> [--severity …] [--limit N]` | Keyword search in NVD |
  618. | `compare <CVE> <CVE> … [--json]` | Compare 2–10 CVEs: patch order with the reason for each place, and what they have in common |
  619. | `epss <CVE,CVE,…>` | EPSS scores |
  620. | `kev <CVE>` | CISA KEV status, and VulnCheck KEV with a token |
  621. | `kev-search [--vendor] [--product] [--keyword] [--ransomware-only] [--added-since] [--due-before] [--epss]` | Filter the CISA KEV catalog, newest first |
  622. | `landscape [--days N] [--markdown]` | Threat briefing for the last 1–30 days |
  623. | `watch add vendor\|product\|package\|cve …` | Watch a vendor, product, package or CVE |
  624. | `watch list\|remove\|refresh\|changes` | List items, stop watching, refresh, or show recorded changes |
  625. | `product <name> [--vendor] [--version] [--cpe]` | CVEs affecting a product, KEV-listed first |
  626. | `fix <package> <ecosystem> <version>` | The version that fixes every known vulnerability |
  627. | `remediate <CVE> [--distro …] [--package …] [--product …] [--apply <option id>]… [--record] [--runbook [--ai] [--out file.md]] [--json]` | The remediation plan for what you run: fixes, workarounds and mitigations in order, each with its sources. `--apply` shows what the options would change, `--record` saves the choice as a decision, `--runbook` writes the steps |
  628. | `scan deps <path>… [--format text\|json\|sarif\|markdown\|cyclonedx] [--fail-on …] [--fail-on-severity …]` | Scan files or folders (`-` for stdin); findings ranked by risk; `cyclonedx` writes an SBOM with the findings in it. See [CI and automation](#ci-and-automation) |
  629. | `scan-deps <file> [--sarif]` | Scan one lockfile, SBOM, manifest or package list; prints JSON |
  630. | `scan exposure <domain\|ip>… --authorized [--no-subdomains] [--max-ips N] [--list-only]` | Passive public exposure: certificate logs, DNS, and what Shodan or InternetDB has seen on your addresses |
  631. | `scan image <file or folder> [--platform os/arch] [--list-only]` | A container image (`docker save`, OCI archive or folder): OS packages and lockfiles, checked with OSV |
  632. | `scan host [--list-only] [--lookup-unknown] [--max-products N]` | This computer's OS and installed apps, checked with NVD; `--list-only` contacts nothing |
  633. | `scans list\|show <id\|latest>\|forget <key>` | Saved scans (`scan … --save` and the app), each with what changed since the scan before |
  634. | `cvss <vector>` | Score a CVSS vector |
  635. | `quick-find <anything>` | Detect the input type and run the right lookup |
  636. | `ip <address>`, `hash <hash>`, `domain <name>` | Threat intelligence for an IP, file hash or domain |
  637. | `password-check` | Check a password against HIBP Pwned Passwords |
  638. | `ask "<question>" [--model …] [--provider …]` | Ask a question; `--provider openai\|anthropic\|gemini\|custom` uses a remote model |
  639. | `ai list\|set\|models\|use\|remove-key\|test` | Configure remote AI for `ask`: OpenAI, Anthropic, Gemini or any OpenAI-compatible endpoint such as LiteLLM; `models` lists what your key can use (see [Remote providers](#remote-providers-opt-in)) |
  640. | `models list\|download\|remove`, `llama status\|install\|install-mlx` | Set up the local model and its runtime (`install-mlx` on Apple silicon) |
  641. | `keys list\|links\|set\|remove` | Manage API keys (see [Terminal configuration](#terminal-configuration)) |
  642. | `sources [--probe]`, `health`, `clear-cache` | Data sources, health check, clear the lookup cache |
  643. | `tls-check [host…] [--json]` | For each host (default: the core sources): the proxy used, the CA store read, the certificate chain presented and whether it verifies. See [HTTPS, proxies and private CAs](#https-proxies-and-private-cas) |
  644. | `mcp`, `serve` | Start the MCP server over stdio, or the HTTP server (MCP and REST) |
  645. | `fleet serve\|enroll\|hosts\|host\|cve\|events\|check\|token\|revoke\|forget` | [Fleet mode](docs/fleet.md): run the central server that agents report to, enroll machines, and see which machines a CVE is on |
  646. `triage`, `fix`, `remediate`, `product`, `kev-search`, `landscape`, `watch`,
  647. `sources` and `ask` print readable text; add `--json` for JSON. `lookup`,
  648. `epss` and `kev` print text, and the other lookup commands print JSON, ready
  649. for `jq`. Run `cve-intel <command> --help` for every option.
  650. #### Recipes
  651. | Situation | Command |
  652. |---|---|
  653. | What happened this week? | `cve-intel landscape --days 7` (add `--markdown` for a ticket) |
  654. | Watch Fortinet and see what changed | `cve-intel watch add vendor fortinet` then `cve-intel watch refresh` and `cve-intel watch changes --unread` |
  655. | A new CVE is in the news. Do we drop everything? | `cve-intel triage CVE-2024-3400 --depth deep` |
  656. | A scanner reported 40 findings. Which come first? | `cve-intel compare CVE-… CVE-…` |
  657. | Which version of a library fixes everything? | `cve-intel fix lodash npm 4.17.15` |
  658. | How do I fix this CVE on what I run? | `cve-intel remediate CVE-2024-3400 --product "PAN-OS@10.2.8"` (or `--distro ubuntu:jammy --package name@version`, or `--package npm:name@version`) |
  659. | What is known about the Tomcat version we run? | `cve-intel product tomcat --vendor apache --version 9.0.80` |
  660. | Is anything from a vendor being exploited right now? | `cve-intel kev-search --vendor fortinet --added-since 2026-01-01` |
  661. | An unknown IP appears in firewall or login logs. | `cve-intel ip 185.220.101.34` |
  662. | EDR flags an unknown file. | `cve-intel hash <sha256>` checks MalwareBazaar, VirusTotal and CIRCL hashlookup. |
  663. | Checking dependencies before a release. | `cve-intel scan deps .` |
  664. ### CI and automation
  665. Fail a CI build on the findings that matter. `scan deps` takes files and
  666. folders (searched like the app does), ranks the findings by risk and exits
  667. with code 3 when a gate trips:
  668. ```bash
  669. cve-intel scan deps . --fail-on kev # anything in CISA KEV
  670. cve-intel scan deps . --fail-on exploited # in KEV, or a public or weaponized exploit
  671. cve-intel scan deps . --fail-on high # risk HIGH or CRITICAL
  672. cve-intel scan deps . --fail-on-severity critical # the advisory's own rating, ignoring risk
  673. ```
  674. | Exit code | Meaning |
  675. |---|---|
  676. | 0 | No gate tripped |
  677. | 1 | An error; or `--strict` and something could not be checked; or a KEV-based gate when CISA KEV could not be loaded |
  678. | 2 | Wrong command-line arguments |
  679. | 3 | A gate tripped (change it with `--exit-code N`) |
  680. The risk score weighs exploitation, so a CVSS 9.8 that nobody exploits can
  681. rank LOW. Use `--fail-on-severity` to gate on advisory severity instead, or
  682. both. Without `--strict`, anything that could not be checked (a package OSV
  683. rejected, EPSS or the exploit indexes unreachable) is reported on stderr and
  684. does not fail the build. `--enrich basic` skips the exploit indexes (about
  685. 50 MB, downloaded once a day); on CI runners, cache `CACHE_DB_PATH` to avoid
  686. fetching them on every run.
  687. `--format markdown` writes a summary table for `$GITHUB_STEP_SUMMARY` or a
  688. pull request comment, `--format json` the full result for `jq`,
  689. `--format cyclonedx` a CycloneDX 1.5 SBOM of everything scanned with each
  690. finding, its risk signals and your decisions attached (for Dependency-Track,
  691. a release artifact or an auditor; it works for `scan image` and `scan host`
  692. too), and `-o FILE` writes to a file:
  693. ```bash
  694. cve-intel scan deps . --format markdown >> "$GITHUB_STEP_SUMMARY"
  695. cve-intel scan deps requirements.txt --format json | jq '.summary'
  696. cve-intel scan deps . --format cyclonedx -o sbom.cdx.json
  697. ```
  698. Show the findings in GitHub code scanning (the **Security** tab and pull
  699. request annotations) by uploading SARIF:
  700. ```yaml
  701. # .github/workflows/dependencies.yml
  702. on: [push, pull_request]
  703. permissions:
  704. contents: read
  705. security-events: write
  706. jobs:
  707. scan:
  708. runs-on: ubuntu-latest
  709. steps:
  710. - uses: actions/checkout@v4
  711. - name: Scan the dependencies with CVE Intel
  712. run: cve-intel scan deps . --format sarif -o cve-intel.sarif
  713. - uses: github/codeql-action/upload-sarif@v3
  714. with:
  715. sarif_file: cve-intel.sarif
  716. category: cve-intel
  717. ```
  718. This assumes `cve-intel` is on the runner's `PATH` (for example the release
  719. AppImage, or `cargo build --release -p cve-server`). Each vulnerability is a
  720. rule with an OSV link and a `security-severity`. Each result points at the
  721. file and line that declares the package, names the advisory and the fixed
  722. version, and carries the risk score, KEV status and EPSS. Findings in KEV or
  723. with HIGH or CRITICAL risk are errors whatever the advisory's own rating.
  724. Get a machine-readable verdict for a ticket or dashboard:
  725. ```bash
  726. cve-intel triage CVE-2024-3400 --depth deep --json | jq '.risk | {risk_score, risk_label, urgency}'
  727. ```
  728. ### REST API
  729. `cve-intel serve` starts an HTTP server on `127.0.0.1:8000` (set `HOST` and
  730. `PORT` to change it, and see [Serving HTTPS](#serving-https) for TLS) with
  731. the REST API under `/api/v1` and MCP at `/mcp`.
  732. The server the desktop app starts from **MCP Server** serves the same
  733. endpoints.
  734. | Method and path | Returns |
  735. |---|---|
  736. | `GET /health` | Version, scoring version, NVD and KEV status (with OAuth on, just `{"status":"ok"}`) |
  737. | `GET /api/v1/health` | The same report, always in full |
  738. | `GET /api/v1/triage/{cve_id}?depth=quick\|standard\|deep` | The full triage report |
  739. | `GET /api/v1/cves/{cve_id}` | The NVD and CVE.org record |
  740. | `GET /api/v1/remediation/{cve_id}?distro=…&package=…&product=…&exposure=…` | The remediation plan, for what you run when a context is given |
  741. | `GET /api/v1/search?q=…&severity=…&limit=…` | NVD keyword search |
  742. | `POST /api/v1/quick-find` with `{"input": "…"}` | Auto-detected lookup |
  743. | `GET /api/v1/kev` | CISA KEV catalog summary |
  744. | `GET /api/v1/landscape?days=7` | The threat-landscape briefing (`format=markdown` for Markdown) |
  745. | `GET /api/v1/watch` | The local watchlist (read-only) |
  746. | `GET /api/v1/watch/changes?unread_only=true` | Recorded watchlist changes (read-only) |
  747. | `GET /api/v1/scans`, `GET /api/v1/scans/{id}`, `GET /api/v1/scans/findings?min_risk=HIGH` | Saved scans and their findings (read-only; only on the server the app starts) |
  748. | `GET /api/v1/tools` | The tools with their input schemas: 42 on the server the app starts, 37 from `cve-intel serve`, which leaves out the tools that read your saved scans and decisions |
  749. | `POST /api/v1/tools/{name}` with the tool's arguments | Runs any tool |
  750. | `GET /api/v1/sources`, `POST /api/v1/sources/probe` | The source catalog, and live connectivity checks |
  751. | `DELETE /api/v1/cache` | Clears the lookup cache |
  752. | `GET /api/v1/models` | The local model catalog |
  753. | `POST /api/v1/ask` with `{"question": "…"}` | An Ask answer from the local model |
  754. | `POST /mcp` | MCP over streamable HTTP |
  755. Errors return HTTP 400 with `{"error": "…"}`.
  756. ```bash
  757. curl -s "http://127.0.0.1:8000/api/v1/triage/CVE-2024-3400?depth=deep" | jq '.risk'
  758. curl -s -X POST http://127.0.0.1:8000/api/v1/tools/find_fixed_version \
  759. -H 'content-type: application/json' \
  760. -d '{"package": "lodash", "ecosystem": "npm", "version": "4.17.15"}'
  761. curl -s -X POST http://127.0.0.1:8000/api/v1/tools/get_attack_mapping \
  762. -H 'content-type: application/json' -d '{"cve_id": "CVE-2024-34102"}'
  763. ```
  764. - `/api/v1/ask` only uses the local model (or `CVE_LLM_BASE_URL`), never a
  765. remote provider, so a caller cannot spend your remote-AI key.
  766. - By default `cve-intel serve` has no authentication and allows
  767. cross-origin requests, so browser front ends can call it. That also means
  768. any web page open in your browser can reach it while it runs, so by
  769. default it only binds to loopback. To serve other machines, turn on OAuth
  770. (see [Securing the HTTP server](#securing-the-http-server)). The server
  771. the desktop app starts refuses web pages instead (see below).
  772. ### Securing the HTTP server
  773. To run `cve-intel serve` for a team or in production, put it behind your
  774. organisation's identity provider. The server then acts as an **OAuth 2.1
  775. resource server**: every request needs an access token except `/health`
  776. and the discovery document, and MCP clients (Claude Code, Cursor, VS Code, …)
  777. discover where to log in from the server itself
  778. ([RFC 9728](https://www.rfc-editor.org/rfc/rfc9728) protected-resource
  779. metadata, as the MCP authorization spec requires). `/health` stays open for
  780. load balancers but only answers `{"status":"ok"}`; the full health report
  781. (NVD and KEV status, which API keys are set) moves to `GET /api/v1/health`,
  782. which needs a token. Each accepted request is logged with the user and
  783. client it came from, and each MCP tool call with the tool's name.
  784. The server never issues tokens, never sees a password and needs no client
  785. secret. It checks each token's signature against the issuer's published
  786. keys (JWKS, found through OIDC discovery and cached), and its `iss`, `aud`,
  787. `exp`, `nbf`, `iat`, `sub` and optional scopes. Only asymmetric algorithms
  788. (RS256/384/512, PS256/384/512, ES256, ES384, EdDSA) are accepted, and a
  789. token with a `crit` header is refused. So is an OpenID Connect ID token:
  790. one with a Keycloak `typ` other than `Bearer`, a Cognito `token_use` other
  791. than `access`, or a `nonce`, `at_hash`, `c_hash` or `s_hash` claim. The
  792. check relies on those markers alone, so an ID token that has none of them
  793. gets in if its `aud` matches. That can only happen when the expected
  794. audience is the ID of a client that users sign in to, which is easy to set
  795. up by accident with Entra ID (see the Entra note below).
  796. | Variable | Meaning |
  797. |---|---|
  798. | `CVE_INTEL_OAUTH_ISSUER` | The authorization server's issuer, exactly as in its tokens' `iss` (keep or omit a trailing slash to match). Setting it turns OAuth on. |
  799. | `CVE_INTEL_OAUTH_RESOURCE` | The public URL of the MCP endpoint, e.g. `https://cve.example.com/mcp`. Required with the issuer. |
  800. | `CVE_INTEL_OAUTH_AUDIENCE` | Expected `aud`, comma-separated for several. Default: the resource URL. |
  801. | `CVE_INTEL_OAUTH_REQUIRED_SCOPES` | Scopes every token must carry (`403 insufficient_scope` otherwise), separated by spaces or commas. Default: none. Set but empty (`""`, `,`) is an error. |
  802. | `CVE_INTEL_OAUTH_SCOPES_SUPPORTED` | Scopes advertised to clients, separated by spaces or commas. Default: `openid` plus the required scopes. |
  803. | `CVE_INTEL_OAUTH_REQUIRED_ROLES` | Who may use the server: roles or groups, separated by commas (a name may contain spaces, e.g. `CVE Admin,/Security Team`); a user needs **any one** of them. Default: none, so any valid token gets in. Set but empty (`""`, `,`) is an error. |
  804. | `CVE_INTEL_OAUTH_JWKS_URI` | Use this JWKS instead of discovering it. |
  805. | `CVE_INTEL_OAUTH_LEEWAY_SECS` | Clock-skew allowance for `exp`, `nbf` and `iat`, 0–600 seconds. Default `60`. |
  806. | `CVE_INTEL_ALLOW_UNAUTHENTICATED` | `1` to serve a non-loopback `HOST` **without** OAuth, for example behind your own authenticating proxy. |
  807. Without OAuth, `cve-intel serve` refuses to start on a non-loopback address
  808. (including Docker's `0.0.0.0`) unless `CVE_INTEL_ALLOW_UNAUTHENTICATED=1` is
  809. set. At startup with OAuth on, it fetches the issuer's discovery document
  810. and keys, and refuses to start if they can't be loaded, if none of the keys
  811. can verify tokens, or if the issuer in the discovery document doesn't match
  812. `CVE_INTEL_OAUTH_ISSUER` exactly. Keys the server doesn't understand (an
  813. ES512, ECDH-ES or ML-DSA key, say) are skipped. The keys are refetched every
  814. hour and when a token names an unknown key; if the issuer moves its keys,
  815. discovery runs again. While the issuer is unreachable, the keys already
  816. held keep working for up to four hours; after that, requests get `503`
  817. until it is back.
  818. Each connection must send a complete request header within 10 seconds of
  819. connecting (or of its previous response), and the server holds at most 512
  820. connections at once, counting those still in the TLS handshake. At the
  821. limit it stops accepting until one closes, and logs a warning at most once
  822. a minute. The connection limit is fixed.
  823. The same 10 seconds also close an idle keep-alive connection. A load
  824. balancer that keeps idle connections to the server open longer (AWS ALB
  825. does for 60 seconds by default) can send a request on one just as the
  826. server closes it, and answer that request with a `502`. Behind one, set
  827. `CVE_INTEL_HEADER_TIMEOUT_SECS` (1–3600 seconds) above the load balancer's
  828. idle timeout, e.g. `65` for an ALB. A longer timeout also gives a client
  829. that never finishes its header longer to hold a connection, so raise it
  830. only when a proxy or load balancer stands between the server and its
  831. clients. `serve` prints the value at startup when it isn't the default.
  832. **Keycloak example.** With a realm `acme` at `https://sso.example.com` and
  833. the server at `https://cve.example.com/mcp`:
  834. 1. Create a client for the resource server, e.g. `cve-intel` (no secret
  835. needed; it only represents the audience). Create a client scope, e.g.
  836. `cve-intel-audience`, with an **Audience** mapper whose *Included Custom
  837. Audience* is exactly `https://cve.example.com/mcp`. Without it,
  838. Keycloak's tokens carry `aud: account` and are rejected.
  839. 2. Let your MCP clients register (dynamic client registration) or create a
  840. public client for them with that scope available.
  841. 3. Configure the server:
  842. ```bash
  843. export CVE_INTEL_OAUTH_ISSUER=https://sso.example.com/realms/acme # no trailing slash
  844. export CVE_INTEL_OAUTH_RESOURCE=https://cve.example.com/mcp
  845. export CVE_INTEL_OAUTH_SCOPES_SUPPORTED="openid offline_access profile cve-intel-audience"
  846. HOST=0.0.0.0 cve-intel serve
  847. ```
  848. Or put the three settings in `.env` (without `export` and without the
  849. comment), since `cve-intel` reads it at startup.
  850. 4. Check it:
  851. ```bash
  852. curl -s https://cve.example.com/.well-known/oauth-protected-resource/mcp
  853. curl -si -X POST https://cve.example.com/mcp -d '{}' | grep -i www-authenticate
  854. claude mcp add --transport http cve-intel https://cve.example.com/mcp # then /mcp to log in
  855. ```
  856. **Troubleshooting.** `cve-intel serve` logs each rejected token's reason
  857. (`wrong audience`, `token expired`, …). `cve-intel --debug serve` logs the
  858. whole exchange: its discovery and JWKS fetches (URLs, status, the
  859. advertised issuer and endpoints, each key's id, type and algorithm), the
  860. metadata it served, and for every request the client, the token's header
  861. and standard claims (`iss`, `aud`, `sub`, `azp`, `scope`, times; control
  862. characters and bidi overrides shown as `\u{..}`), what the server
  863. expected, each check's result and the challenge it sent back. Tokens appear
  864. only as a fingerprint (`tok:` and 12 hex digits of their SHA-256) so one
  865. token can be followed across lines; the token, its signature and the
  866. `Authorization` header are never logged. A Keycloak token with
  867. `aud="account"` in that log means the audience mapper or the
  868. `<client>-audience` scope is missing, and `not an access token` means the
  869. client sent its ID token instead. Without `--debug`, only cve-intel's own
  870. lines appear; messages from the TLS, HTTP and proxy libraries (a scanner's
  871. illegal SNI, say) need `--debug`.
  872. Notes for other providers: Entra ID puts scopes in `scp` (accepted) and
  873. uses `api://<app-id>` or the app ID as audience, so set
  874. `CVE_INTEL_OAUTH_AUDIENCE`; Auth0's issuer ends in `/`. Serve the public
  875. URL over HTTPS, from cve-intel itself ([Serving HTTPS](#serving-https)) or
  876. from a proxy in front of it; the URLs in the
  877. metadata and the `WWW-Authenticate` challenge come from
  878. `CVE_INTEL_OAUTH_RESOURCE`, never from request headers. If the proxy does
  879. its own authentication, exempt `/mcp`, `/api/v1` and
  880. `/.well-known/oauth-protected-resource*` so its challenge doesn't replace
  881. the server's.
  882. **Entra ID and ID tokens.** An Entra ID token's `aud` is the client ID of
  883. the app the user signed in to, and one from the code exchange carries no
  884. `at_hash` or `c_hash`, and no `nonce` unless the client sent one. If the
  885. MCP client signs in with the API's own app registration, so that
  886. `CVE_INTEL_OAUTH_AUDIENCE` is that client's ID, its ID token passes every
  887. check. Prevent it in one of two ways:
  888. - Give the API its own app registration, separate from the clients'. A
  889. client's ID token then carries the client's ID, not the API's, and fails
  890. the audience check.
  891. - Set `CVE_INTEL_OAUTH_REQUIRED_SCOPES` to a scope the API exposes (e.g.
  892. `access_as_user`). Entra puts `scp` only in access tokens issued to
  893. users, so an ID token never has it. This also refuses app-only
  894. (client-credentials) tokens, which carry `roles` instead of `scp`.
  895. `CVE_INTEL_OAUTH_REQUIRED_ROLES` doesn't help here, because Entra puts the
  896. user's app roles in ID tokens too.
  897. **Who may use it.** Without `CVE_INTEL_OAUTH_REQUIRED_ROLES`, any valid
  898. token for this server gets in. Set it to restrict the server to users who
  899. hold one of the listed roles or groups:
  900. ```bash
  901. CVE_INTEL_OAUTH_REQUIRED_ROLES=cve-intel-user,secops
  902. ```
  903. Separate entries with commas only, because role and group names may
  904. contain spaces (`CVE Admin,/Security Team` is two entries).
  905. Each entry is matched exactly (case-sensitive) against the token's:
  906. | Claim | Write it as | Comes from |
  907. |---|---|---|
  908. | `realm_access.roles` | `cve-intel-user` | Keycloak realm roles (the `roles` client scope) |
  909. | `resource_access.<client>.roles` | `cve-intel:admin` (`client:role`) | Keycloak client roles; only that client's roles count |
  910. | `groups` | `secops` or `/org/secops` | Keycloak's group membership mapper (the `microprofile-jwt` scope, or a mapper you add), Okta; a top-level group's leading `/` is optional, a nested group needs its full path |
  911. | `roles` | `CveIntel.User` | Entra ID app roles |
  912. A user who authenticates but holds none of them gets a plain `403
  913. access_denied`, with no bearer challenge (logging in again wouldn't help)
  914. and without the list of allowed roles. The server logs who was refused,
  915. and `--debug` shows which role and group claims the token carried. For
  916. Keycloak, make sure the clients your users connect with receive the
  917. `roles` scope (or your group mapper), or the claims won't be in the token;
  918. adding `roles` to `CVE_INTEL_OAUTH_SCOPES_SUPPORTED` asks clients for it.
  919. Saved scans and decisions are never served by `cve-intel serve`, so a
  920. permitted user can use every tool the server offers.
  921. #### Serving HTTPS
  922. OAuth providers and MCP clients expect the server's URL to be `https://`.
  923. `cve-intel serve` can serve HTTPS itself, with no reverse proxy, given a
  924. certificate and its private key:
  925. | Variable | Meaning |
  926. |---|---|
  927. | `CVE_INTEL_TLS_CERT` | The certificate chain in PEM: the server's certificate first, then any intermediates (certbot's `fullchain.pem`). |
  928. | `CVE_INTEL_TLS_KEY` | Its private key in PEM, unencrypted (certbot's `privkey.pem`): RSA of 2048 bits or more, ECDSA P-256 or P-384, or Ed25519. |
  929. ```bash
  930. CVE_INTEL_TLS_CERT=/etc/cve-intel/tls/fullchain.pem \
  931. CVE_INTEL_TLS_KEY=/etc/cve-intel/tls/privkey.pem \
  932. HOST=0.0.0.0 PORT=8443 cve-intel serve
  933. # [cve-intel] MCP at https://0.0.0.0:8443/mcp · …
  934. # [cve-intel] HTTPS on: certificate "/etc/cve-intel/tls/fullchain.pem" (SHA-256 3F:A2:…), reloaded when it changes
  935. ```
  936. - Set both or neither. If only one is set, a file can't be read, or the
  937. key doesn't belong to the certificate, `serve` refuses to start rather
  938. than fall back to plain HTTP. With TLS on, the port speaks only HTTPS.
  939. - The files are read again every minute, and a renewed certificate is used
  940. for new connections without a restart; the log shows its fingerprint
  941. (compare with `openssl x509 -noout -fingerprint -sha256 -in <file>`). If
  942. the new files don't load, for example while the certificate has been
  943. replaced but the key not yet, the server logs why once and keeps serving
  944. the previous certificate.
  945. - HTTPS protects the connection, not who may use it: a non-loopback `HOST`
  946. still needs OAuth (or `CVE_INTEL_ALLOW_UNAUTHENTICATED=1`).
  947. - Clients must trust the certificate, so use a public CA such as Let's
  948. Encrypt, or your organisation's CA installed on each client (Node-based
  949. clients also read `NODE_EXTRA_CA_CERTS`). A self-signed certificate
  950. works only where each client is set up to trust it. The certificate's
  951. names must include the host in `CVE_INTEL_OAUTH_RESOURCE`.
  952. - cve-intel doesn't obtain certificates itself. Get them with certbot (a
  953. standalone or DNS challenge), your CA or cert-manager. The user running
  954. `cve-intel` must be able to read the key, and certbot's keys are readable
  955. only by root, so copy them on each renewal, for example with
  956. `certbot certonly … --deploy-hook 'install -m 0640 -g cve-intel "$RENEWED_LINEAGE/fullchain.pem" "$RENEWED_LINEAGE/privkey.pem" /etc/cve-intel/tls/'`.
  957. - Ports below 1024 need privileges. Use a high port with a port forward
  958. (Docker: `-p 443:8000`), or on Linux grant the binary
  959. `cap_net_bind_service` with `setcap`.
  960. - TLS 1.2 and 1.3 only (rustls). `cve-intel --debug serve` logs each failed
  961. handshake and its reason, such as a client that doesn't trust the
  962. certificate.
  963. ### MCP server setup
  964. MCP is an open protocol and works with many clients, not only Claude. The
  965. easiest setup is the **MCP Server** view in the desktop app (see
  966. [MCP server for AI agents](#mcp-server-for-ai-agents)). To configure it by
  967. hand:
  968. ```bash
  969. # HTTP: connect to the server started from the app, or from `cve-intel serve`
  970. claude mcp add --transport http cve-intel http://127.0.0.1:8000/mcp
  971. # stdio: the agent launches its own copy (macOS app path shown)
  972. claude mcp add cve-intel -- "/Applications/CVE Intel.app/Contents/MacOS/cve-intel" mcp
  973. # stdio, using the headless binary
  974. claude mcp add cve-intel -- /absolute/path/to/target/release/cve-intel mcp
  975. ```
  976. Ready-made configs for Claude Code, Claude Desktop, Cursor, VS Code and
  977. other clients are in [`examples/mcp/`](examples/mcp/).
  978. The server exposes 42 tools, resources (`kev://catalog`,
  979. `epss://scores/{cve_id}`, `manifest://tool-hash`) and prompts
  980. (`patch_decision`, `compare_and_prioritize`, `dependency_triage`,
  981. `remediation_runbook`). It never
  982. registers a sampling handler or issues `sampling/createMessage`. Older tool
  983. names that duplicated other tools (`prioritize_cves`,
  984. `check_exploit_availability`, `scan_container_packages`,
  985. `scan_repo_secrets`) are no longer listed but still work. The watchlist
  986. tools (`list_watchlist`, `get_watchlist_changes`) are read-only: agents
  987. cannot add, remove or refresh items. The scan history tools (`list_scans`,
  988. `get_scan_findings`) are read-only too, and only answer local callers: MCP
  989. over stdio and the server the app starts. `cve-intel serve` accepts
  990. requests from any web page, so it does not serve saved scans.
  991. The server started from the app always binds to `127.0.0.1` and checks the
  992. `Host`, `Origin` and `Sec-Fetch-Site` headers to refuse cross-site browser
  993. requests, which protects against DNS rebinding. It has no sign-in, so any
  994. program on the same computer can use it with your saved keys. Stop it when
  995. you don't need it.
  996. ### Terminal configuration
  997. Manage API keys without the app:
  998. ```bash
  999. cve-intel keys list # which keys are set and where from (never prints values)
  1000. cve-intel keys links # where to get each key
  1001. cve-intel keys set SHODAN_KEY # hidden prompt, keeps the key out of shell history
  1002. echo "$KEY" | cve-intel keys set SHODAN_KEY # or pipe it in from a script
  1003. cve-intel keys remove SHODAN_KEY
  1004. ```
  1005. For headless use, copy [`.env.example`](.env.example) to `.env`, or set
  1006. environment variables. Everything works without keys, with fewer results. A
  1007. key saved in the app takes precedence over an environment variable.
  1008. In `.env`, one `KEY=value` per line, optionally after `export`. Quote a
  1009. value that contains ` #`: unquoted, a `#` after whitespace starts a comment
  1010. (`PASS=#abc` keeps its `#`). Inside double quotes, `\"` and `\\` stand for
  1011. `"` and `\`; single quotes are literal. If a line is ignored, or text is cut
  1012. from a value (a comment, or text after the closing quote), `cve-intel`
  1013. prints a warning naming the line and key, never the value.
  1014. | Purpose | Variables |
  1015. |---|---|
  1016. | Data sources | `NVD_API_KEY`, `GITHUB_TOKEN`, `VULNCHECK_TOKEN`, `ABUSEIPDB_KEY`, `VIRUSTOTAL_KEY`, `SHODAN_KEY`, `GREYNOISE_API_KEY`, `URLSCAN_KEY`, `CIRCL_PDNS_USER`, `CIRCL_PDNS_PASS`, `ABUSECH_AUTH_KEY` |
  1017. | Remote AI | `OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `GEMINI_API_KEY`, `CUSTOM_LLM_BASE_URL`, `CUSTOM_LLM_API_KEY` |
  1018. | Local AI | `CVE_LLM_BASE_URL`, `CVE_LLM_MODEL`, `LLAMA_SERVER_BIN`, `MLX_LM_SERVER_BIN` |
  1019. | Storage and server | `CACHE_DB_PATH`, `CVE_INTEL_HOME`, `HOST`, `PORT`, `CVE_INTEL_HEADER_TIMEOUT_SECS`, `REQUEST_TIMEOUT` |
  1020. #### HTTPS, proxies and private CAs
  1021. Certificates are checked by the operating system's verifier, so a CA you
  1022. installed in the OS is trusted, including the CA of a TLS-inspecting proxy:
  1023. the macOS keychains, the Windows certificate store, or on Linux the system
  1024. bundle (`update-ca-trust` on RHEL/Amazon Linux, `update-ca-certificates` on
  1025. Debian/Ubuntu). `SSL_CERT_FILE` / `SSL_CERT_DIR` replace the Linux bundle.
  1026. `HTTPS_PROXY`, `ALL_PROXY` and `NO_PROXY` are honoured.
  1027. If a lookup fails with `invalid peer certificate: UnknownIssuer`, run
  1028. `cve-intel tls-check`. It shows the issuer at the top of the chain the
  1029. server sent, which is the CA missing from your store. Add `--debug` to any
  1030. command to log TLS and connection details to stderr.
  1031. #### Use your own model server
  1032. Set `CVE_LLM_BASE_URL` to any OpenAI-compatible endpoint, such as Ollama
  1033. (`http://127.0.0.1:11434/v1`), LM Studio or your own llama-server, and
  1034. optionally `CVE_LLM_MODEL` to the model name. Ask then uses it instead of
  1035. the built-in runtime. To use a `llama-server` or `mlx_lm.server` you
  1036. installed yourself, set `LLAMA_SERVER_BIN` or `MLX_LM_SERVER_BIN` to its
  1037. path. To use a remote provider from the CLI, configure it with `cve-intel ai`
  1038. (see [Remote providers](#remote-providers-opt-in)), then
  1039. `cve-intel ask "Should we patch CVE-2024-3400?" --provider anthropic`.
  1040. ### Build from source and Docker
  1041. ```bash
  1042. # Desktop app (debug)
  1043. cargo build -p cve-intel-desktop
  1044. ./target/debug/cve-intel-desktop
  1045. # Installable bundle (.dmg / .exe / .AppImage)
  1046. cargo install tauri-cli --version "^2" --locked
  1047. cd apps/desktop/src-tauri && cargo tauri build
  1048. # Headless binary only (servers, Docker)
  1049. cargo build --release -p cve-server
  1050. # The fleet agent, small and stripped (target/agent/cve-intel-agent)
  1051. cargo build --profile agent -p cve-agent
  1052. ```
  1053. The same code also builds as a separate application, with the working name
  1054. CVE Intel Fleet, that installs next to CVE Intel with its own name and data
  1055. folder (`~/.cve-intel-fleet`). Add `--features edition-fleet` to the
  1056. commands above, and for the installable bundle
  1057. `--config tauri.fleet.conf.json` as well. See
  1058. [CONTRIBUTING.md](CONTRIBUTING.md#the-fleet-edition-the-same-code-as-a-separate-application).
  1059. Docker runs the headless MCP server over HTTP, plus the REST API. The
  1060. container listens on `0.0.0.0`, so it needs OAuth, or an explicit opt-out
  1061. when only this machine can reach it:
  1062. ```bash
  1063. docker build -t cve-intel .
  1064. # Production: OAuth settings in .env (see "Securing the HTTP server")
  1065. docker run --rm -p 8000:8000 --env-file .env cve-intel
  1066. # Local only: published on the host's loopback, no sign-in
  1067. docker run --rm -p 127.0.0.1:8000:8000 -e CVE_INTEL_ALLOW_UNAUTHENTICATED=1 --env-file .env cve-intel
  1068. # HTTPS from the container itself (see "Serving HTTPS")
  1069. docker run --rm -p 443:8000 --env-file .env \
  1070. -v /etc/cve-intel/tls:/tls:ro --group-add "$(getent group cve-intel | cut -d: -f3)" \
  1071. -e CVE_INTEL_TLS_CERT=/tls/fullchain.pem -e CVE_INTEL_TLS_KEY=/tls/privkey.pem cve-intel
  1072. ```
  1073. For HTTPS, mount the directory that holds the certificate and key, not
  1074. the files themselves: a renewal replaces the files, and a single-file
  1075. mount keeps showing the old one. The container runs as an unprivileged
  1076. user, so `--group-add` gives it the group that may read the key.
  1077. To run the desktop app without installing it, use the container below and open http://127.0.0.1:6080 to see it over noVNC:
  1078. ```bash
  1079. docker build -f Dockerfile.desktop -t cve-intel-desktop .
  1080. docker run --rm -p 127.0.0.1:6080:6080 --env-file .env -v cve-intel-data:/home/app cve-intel-desktop
  1081. ```
  1082. - noVNC asks for a password. The container prints a random one to its log at start-up. To pick your own, pass `-e VNC_PASSWORD=...` (VNC only checks the first 8 characters).
  1083. - Keep `127.0.0.1:` in the `-p` flag. Without it, anyone on your network can reach the app.
  1084. ### Architecture
  1085. ```
  1086. crates/cve-core Harness-agnostic engine: the source clients, risk scorer,
  1087. SSVC, CVSS, validators, lockfile/SBOM parsers, SARIF
  1088. and CycloneDX output, SQLite cache, settings DB, activity log and
  1089. watchlist, decisions, landscape briefing, source catalog,
  1090. remediation evidence, plans and runbooks, the CSAF
  1091. reader, llama.cpp and MLX manager, cloud-provider
  1092. settings, the Ask agent and fleet mode (its protocol,
  1093. database, checks and agent logic). No MCP, Tauri or CLI
  1094. types.
  1095. crates/cve-server The cve-intel binary: MCP (stdio and streamable HTTP),
  1096. REST (/api/v1), the CLI and the fleet server, as thin
  1097. adapters over cve-core.
  1098. crates/cve-agent The cve-intel-agent binary: the fleet agent. It links only
  1099. cve-core's inventory readers and fleet client, so it is a
  1100. fraction of the size.
  1101. apps/desktop Tauri app linking cve-core and cve-server natively. The
  1102. same executable runs the GUI, the CLI and the MCP server,
  1103. and can host MCP over HTTP in-process. The model server
  1104. (llama-server, or mlx_lm.server on Apple silicon) is the
  1105. only child process. The Fleet view is ui/fleet.js and
  1106. src-tauri/src/fleet.rs.
  1107. examples/mcp Client configs for common MCP clients.
  1108. examples/fleet systemd units for the fleet server and agent.
  1109. ```
  1110. ### Development and tests
  1111. ```bash
  1112. cargo test -p cve-core -p cve-server -p cve-agent # scorer, SSVC, CVSS, validators, cache, parsers, agent routing, source catalog, fleet
  1113. cargo test -p cve-core -p cve-server -- --ignored # live checks against the real sources (network)
  1114. cargo build --release # cve-intel and the desktop binary
  1115. ```
  1116. See [CONTRIBUTING.md](CONTRIBUTING.md) for the full development setup and
  1117. the checks CI runs.
  1118. ## Citing and attribution
  1119. ### Citing results in a report or ticket
  1120. CVE Intel gathers facts from public sources; it is not the source. When you
  1121. record a decision, cite the **primary source and the retrieval date**, not
  1122. the app or the AI's wording:
  1123. > *Example (illustrative values):* CVE-2021-44228: CVSS 10.0 (NVD); listed in
  1124. > CISA KEV, added 2021-12-10 (CISA); EPSS 0.94 as of 2026-10-18 (FIRST).
  1125. > Retrieved 2026-10-18 with CVE Intel 0.3.0, scoring v1.1.
  1126. - **Scores change.** EPSS updates daily, KEV grows and NVD entries are
  1127. revised. Results are cached for minutes to hours, and up to a day for some
  1128. feeds. The **Databases** view shows each source's cache time.
  1129. - **The risk score is CVE Intel's own.** Cite it as *"CVE Intel risk score
  1130. (scoring v1.1)"* alongside the underlying CVSS, EPSS and KEV values.
  1131. `cve-intel health` and the **System** view show the version.
  1132. - **An AI answer is not evidence.** Local or cloud, a model can be wrong.
  1133. Cite the data shown beside the answer and verify it against the source.
  1134. ### Citing this software
  1135. ```bibtex
  1136. @software{cve_intel,
  1137. title = {CVE Intel: local security intelligence for CVE triage},
  1138. author = {{CVE Intel contributors}},
  1139. year = {2026},
  1140. version = {0.3.0},
  1141. url = {https://github.com/Fazmin/cve-intel},
  1142. license = {MIT OR Apache-2.0},
  1143. note = {Native Rust desktop app, CLI, MCP server and REST API}
  1144. }
  1145. ```
  1146. Use the version you ran (`cve-intel --version`).
  1147. ### Data source terms
  1148. CVE Intel only queries these services. It does not own or republish their
  1149. data, and their terms apply to anything you copy out of the app. Check each
  1150. provider's current terms, especially before commercial use. This is not
  1151. legal advice.
  1152. | Source | Credit and terms |
  1153. |---|---|
  1154. | **NVD** (NIST) | Products using the NVD API should state: *"This product uses data from the NVD API but is not endorsed or certified by the NVD."* See the [terms of use](https://nvd.nist.gov/developers/terms-of-use). |
  1155. | **EPSS** (FIRST) | Attribute FIRST and link [first.org/epss](https://www.first.org/epss/). Daily score files are hosted at [epss.empiricalsecurity.com](https://epss.empiricalsecurity.com). |
  1156. | **CVSS** (FIRST) | [first.org/cvss](https://www.first.org/cvss/) |
  1157. | **CISA KEV** | Credit CISA: [Known Exploited Vulnerabilities Catalog](https://www.cisa.gov/known-exploited-vulnerabilities-catalog). |
  1158. | **ENISA EUVD** | EU Vulnerability Database: [euvd.enisa.europa.eu](https://euvd.enisa.europa.eu/). Subject to ENISA's terms. |
  1159. | **CISA advisories** | RSS feed of CISA alerts and ICS advisories: [cisa.gov](https://www.cisa.gov/news-events/cybersecurity-advisories). |
  1160. | **PoC-in-GitHub** | Repository index by [nomi-sec/PoC-in-GitHub](https://github.com/nomi-sec/PoC-in-GitHub). Repositories are listed, not scanned. |
  1161. | **Project Zero 0day In the Wild** | Google Project Zero's public spreadsheet: [0day.html](https://googleprojectzero.blogspot.com/p/0day.html). |
  1162. | **CIRCL Vulnerability-Lookup** | Sightings from [vulnerability.circl.lu](https://vulnerability.circl.lu/). |
  1163. | **CVE.org** | CVE records are provided under the [CVE Terms of Use](https://www.cve.org/Legal/TermsOfUse). CVE® is a registered trademark of The MITRE Corporation. |
  1164. | **CISA Vulnrichment** | CISA's ADP enrichment (SSVC, CVSS, CWE), released under CC0 1.0: [github.com/cisagov/vulnrichment](https://github.com/cisagov/vulnrichment). |
  1165. | **SSVC** | CERT/CC and CISA: [github.com/CERTCC/SSVC](https://github.com/CERTCC/SSVC) |
  1166. | **VulnCheck KEV, XDB and NVD++** | Shown with VulnCheck's attribution, *"VulnCheck KEV (VulnCheck Known Exploited Vulnerabilities), https://vulncheck.com/kev"*. Use is subject to [VulnCheck's terms](https://vulncheck.com/terms). |
  1167. | **Metasploit Framework** | Module metadata from [rapid7/metasploit-framework](https://github.com/rapid7/metasploit-framework), BSD-3-Clause. |
  1168. | **MITRE ATT&CK®** | Registered trademark of The MITRE Corporation. See the [terms of use](https://attack.mitre.org/resources/terms-of-use/). |
  1169. | **MITRE CAPEC™** | Attack patterns from [capec.mitre.org](https://capec.mitre.org); see the [terms of use](https://capec.mitre.org/about/termsofuse.html). |
  1170. | **CTID Mappings Explorer** | KEV → ATT&CK mappings by the Center for Threat-Informed Defense, Apache-2.0: [mappings-explorer](https://github.com/center-for-threat-informed-defense/mappings-explorer). |
  1171. | **Arch Linux Security Tracker** | Advisory groups published by the Arch Linux Security Team at [security.archlinux.org](https://security.archlinux.org). |
  1172. | **OSV.dev**, **GitHub Advisory Database** | Attribution varies by database; the GitHub Advisory Database is CC BY 4.0. See [osv.dev](https://osv.dev) and [github.com/advisories](https://github.com/advisories). |
  1173. | **abuse.ch** (MalwareBazaar, ThreatFox) | Community data; see [abuse.ch](https://abuse.ch). |
  1174. | Others | VirusTotal, Shodan, AbuseIPDB, GreyNoise, URLScan, CIRCL, Red Hat, Ubuntu, Microsoft MSRC, crt.sh, Exploit-DB, Ransomwhere and HIBP each have their own terms and rate limits. Free tiers are often limited to non-commercial use. |
  1175. **Local AI models** are not part of this project and carry their own
  1176. licences, such as the Llama 3.2 Community License and size-dependent Qwen
  1177. licences. Read the model's page before commercial use. GGUF files come from
  1178. the Hugging Face repositories listed in `crates/cve-core/src/llm.rs` (the
  1179. vendors' own for Qwen, bartowski's builds for Llama), and MLX builds from
  1180. the [mlx-community](https://huggingface.co/mlx-community) conversions. The
  1181. runtimes are [llama.cpp](https://github.com/ggml-org/llama.cpp) (MIT) and,
  1182. on Apple silicon, [mlx-lm](https://github.com/ml-explore/mlx-lm) (MIT).
  1183. ## Community
  1184. - [CONTRIBUTING.md](CONTRIBUTING.md): development setup and pull request guidelines
  1185. - [CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md): community standards
  1186. - [SECURITY.md](SECURITY.md): reporting a vulnerability privately
  1187. ## License
  1188. Copyright 2024-2025 Fazmin Nizam. Licensed under either of
  1189. - Apache License, Version 2.0 ([LICENSE-APACHE](LICENSE-APACHE) or
  1190. <https://www.apache.org/licenses/LICENSE-2.0>)
  1191. - MIT License ([LICENSE-MIT](LICENSE-MIT) or
  1192. <https://opensource.org/licenses/MIT>)
  1193. at your option. See also [NOTICE](NOTICE).
  1194. These licenses cover the CVE Intel code. Data and models the app downloads
  1195. keep their own terms; see [Data source terms](#data-source-terms).
  1196. Unless you explicitly state otherwise, any contribution intentionally
  1197. submitted for inclusion in the work by you, as defined in the Apache-2.0
  1198. license, shall be dual licensed as above, without any additional terms or
  1199. conditions.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.