catude
catude, a terminal UI for Claude Code and Claude Desktop, as slash commands inside Claude Code.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/catude--catude)HTML
<a href="/listings/catude--catude"><img src="/api/listings/catude--catude/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
These steps run a script from the network that findsafeskills does not scan.
Each line downloads the latest release, checks its checksum and puts `catude` and the short `ct` on your PATH. No administrator rights are needed.
# Windows (PowerShell) irm https://raw.githubusercontent.com/catude/catude/main/scripts/install.ps1 | iex
# Linux, macOS, WSL curl -fsSL https://raw.githubusercontent.com/catude/catude/main/scripts/install.sh | sh
# With Go 1.24 or later go install github.com/catude/catude/cmd/catude@latest
Or download the archive for your system from [Releases](https://github.com/catude/catude/releases). The first run offers to install itself, with the answers already selected; `catude uninstall` removes it. Then:
catude why # where your tokens went catude # the terminal app (or: ct)
> Works on Windows 10 and 11, macOS, Linux and WSL. The Linux and macOS builds are **experimental**: tested on every > change, but with little real-world use so far ([Linux notes](docs/linux.md), [Windows notes](docs/windows.md)).
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.
.claude-plugin/marketplace.json · Manifest · 16 lines
- {
- "name": "catude",
- "owner": {
- "name": "catude"
- },
- "description": "catude, a terminal UI for Claude Code and Claude Desktop, as slash commands inside Claude Code.",
- "plugins": [
- {
- "name": "catude",
- "source": "./plugin",
- "description": "Ask Claude about your own Claude setup with catude: where configuration is read, security findings, context use of recent sessions and which model to pick. Needs the catude program on your PATH.",
- "category": "productivity",
- "homepage": "https://github.com/catude/catude"
- }
- ]
- }
README.md · README · 481 lines · 4 findings marked
- <p align="center">
- <picture>
- <source media="(prefers-color-scheme: dark)" srcset="assets/logo-dark.svg">
- <img alt="catude" src="assets/logo-light.svg" width="320">
- </picture>
- </p>
- <h3 align="center">Find out why your Claude Code usage runs out, trim what loads into every session,<br>and see everything that configures Claude Code and Claude Desktop on your machine.</h3>
- <p align="center">
- A terminal app and command line. One binary, no account, no telemetry. Windows, macOS, Linux and WSL.
- </p>
- <p align="center">
- <a href="#install"><b>Install</b></a> ·
- <a href="#what-it-answers">What it answers</a> ·
- <a href="#command-line">Command line</a> ·
- <a href="https://catude.github.io/catude/"><b>Docs</b></a> ·
- <a href="docs/security.md">Security</a> ·
- <a href="docs/comparison.md">Alternatives</a>
- </p>
- <p align="center">
- <a href="https://github.com/catude/catude/actions/workflows/ci.yml"><img alt="Tests" src="https://github.com/catude/catude/actions/workflows/ci.yml/badge.svg"></a>
- <a href="https://github.com/catude/catude/actions/workflows/codeql.yml"><img alt="CodeQL" src="https://github.com/catude/catude/actions/workflows/codeql.yml/badge.svg"></a>
- <a href="https://github.com/catude/catude/releases"><img alt="Latest release" src="https://img.shields.io/github/v/release/catude/catude?color=D97757"></a>
- <a href="#security-you-can-check"><img alt="Signed build provenance" src="https://img.shields.io/badge/releases-signed%20provenance-D97757"></a>
- <img alt="Windows, macOS, Linux, WSL" src="https://img.shields.io/badge/runs%20on-Windows%20%C2%B7%20macOS%20%C2%B7%20Linux%20%C2%B7%20WSL-informational">
- <img alt="No telemetry" src="https://img.shields.io/badge/telemetry-none-success">
- <img alt="10 languages" src="https://img.shields.io/badge/languages-10-blueviolet">
- <img alt="MIT" src="https://img.shields.io/badge/license-MIT-blue">
- </p>
- <p align="center"><img alt="TL;DR: where your tokens went, switches per project, the agent map, work profiles, presets, Discover, config, audit, live sessions" src="assets/tldr.gif" width="900"></p>
- > **TL;DR**
- > - **`catude why`**: where your Claude Code tokens went (cache rebuilt after pauses, subagents, `claude -p`, the context before you type) and what to change.
- > - **Switch things off per project**: `x` turns a skill or MCP server off for one project, with the settings Claude Code documents.
- > - **Agent map**: which subagents run, who launches them, the tools they use and what they cost.
- > - **Work profiles**: one set of skills, agents, MCP servers and plugins per kind of work; `ct switch` changes which one plain `claude` starts with.
- > - **Presets of the best**: start from the published setups of Boris Cherny, Simon Willison, Mitchell Hashimoto, Matt Pocock and 11 more, each with its sources.
- > - **Discover**: 177 plugins, 89 skills, 55 marketplaces and 25 MCP servers, checked at their source; `i` twice installs with the official `claude` command.
- > - **Config**: every documented setting of Claude Code and Claude Desktop, which file wins, and 30 presets that quote their source.
- > - **Audit**: 25 rules for secrets in permission rules, `curl | sh`, hidden Unicode and other risks in what you installed.
- > - **Live**: what every running session is doing, with a notice (or a system notification) when one waits for you.
- > - **One binary**: no account, no telemetry, reads local files only. Windows, macOS, Linux, WSL. 10 languages.
- ## Try it
- <details open>
- <summary><b>Windows</b> (PowerShell)</summary>
- ```powershell
- irm https://raw.githubusercontent.com/catude/catude/main/scripts/install.ps1 | iex
- Note: Runs a downloaded PowerShell script (iex / Invoke-Expression)
- Line 54 on GitHub
- ```
- Then open a new terminal and run `catude why` or `catude`.
- </details>
- <details>
- <summary><b>Linux, macOS, WSL</b> (experimental)</summary>
- ```sh
- curl -fsSL https://raw.githubusercontent.com/catude/catude/main/scripts/install.sh | sh
- Note: Pipes a downloaded script directly into a shell (curl | sh / wget | bash)
- Line 64 on GitHub
- ```
- Then open a new terminal and run `catude why` or `catude`. The Linux and macOS builds are tested on every change but
- have little real-world use so far: [please report what you find](https://github.com/catude/catude/issues/new/choose).
- </details>
- <details>
- <summary><b>With Go</b> (any system)</summary>
- ```sh
- go install github.com/catude/catude/cmd/catude@latest
- ```
- </details>
- Each script downloads the [latest release](https://github.com/catude/catude/releases/latest), checks its checksum and
- puts `catude` and the short `ct` on your PATH, without administrator rights. More options in [Install](#install).
- <p align="center"><img alt="catude: where the tokens went, the agent map, every file Claude reads and the live sessions" src="assets/demo.gif" width="900"></p>
- ## What it answers
- | Question | Where |
- |---|---|
- | **Why did my usage run out this week?** | [`catude why`](#why-your-usage-runs-out), Usage tab |
- | **What loads into every session before I type, and how do I trim it for one project?** | [Context, MCP and Skills tabs](#trim-what-loads-into-every-session) |
- | **Which subagents run, who launches them and what do they cost?** | [Agents tab, `catude agents`](#the-agent-map) |
- | **Did something get worse after an update?** | [`catude health`](#your-own-evidence-when-something-feels-worse) |
- | **Is anything I installed unsafe?** | [Audit tab, `catude audit`](#audit-what-you-installed) |
- | **What is each running session doing right now?** | [Live tab](#live-what-every-session-is-doing) |
- | **How do I keep one setup per kind of work?** | [Work profiles](#work-profiles-one-setup-per-kind-of-work) |
- | **Which file wins when a setting is in three places?** | [Config tab](#every-setting-and-which-one-wins) |
- Everything comes from files already on your machine: Claude Code's transcripts and settings, and Claude Desktop's
- configuration. catude changes nothing unless you press a key that does, backs up every file first, and never shows
- or exports a secret.
- ## Install
- Each line downloads the latest release, checks its checksum and puts `catude` and the short `ct` on your PATH.
- No administrator rights are needed.
- ```powershell
- # Windows (PowerShell)
- irm https://raw.githubusercontent.com/catude/catude/main/scripts/install.ps1 | iex
- Note: Runs a downloaded PowerShell script (iex / Invoke-Expression)
- Line 108 on GitHub
- ```
- ```sh
- # Linux, macOS, WSL
- curl -fsSL https://raw.githubusercontent.com/catude/catude/main/scripts/install.sh | sh
- Note: Pipes a downloaded script directly into a shell (curl | sh / wget | bash)
- Line 113 on GitHub
- ```
- ```sh
- # With Go 1.24 or later
- go install github.com/catude/catude/cmd/catude@latest
- ```
- Or download the archive for your system from [Releases](https://github.com/catude/catude/releases). The first run offers to install itself,
- with the answers already selected; `catude uninstall` removes it. Then:
- ```sh
- catude why # where your tokens went
- catude # the terminal app (or: ct)
- ```
- > Works on Windows 10 and 11, macOS, Linux and WSL. The Linux and macOS builds are **experimental**: tested on every
- > change, but with little real-world use so far ([Linux notes](docs/linux.md), [Windows notes](docs/windows.md)).
- ## Why your usage runs out
- <p align="center"><img alt="catude why: subagents, cache rebuilt after pauses, one expensive conversation, claude -p runs and the context before the first message" src="assets/screenshots/cli-why.png" width="900"></p>
- Counting tokens says *how much*. catude says *where they went*, from the API usage Claude Code records in your
- transcripts:
- - **Cache rebuilt after a pause.** The prompt cache expires after a few minutes without activity; the next message
- sends the whole conversation again. catude counts those rebuilds and what they cost.
- - **Subagents.** Each one starts its own context and re-reads files.
- - **Headless runs.** Scripts, hooks and tools that start `claude -p` count against the same limits.
- - **The context before your first message**: system prompt, tools, MCP servers, `CLAUDE.md` and skill descriptions.
- - **The most expensive conversations**, and **whether a new Claude Code version** uses more tokens per response.
- A conversation resumed or forked into another folder copies its history; catude counts those responses once. The
- Usage tab shows the same next to your days, models and projects, and, with
- [claude-swap](https://github.com/realiti4/claude-swap), when each plan window resets and whether your pace lasts.
- <p align="center"><img alt="Usage tab: where the tokens went over 7, 14 and 30 days" src="assets/quota.gif" width="900"></p>
- ## Trim what loads into every session
- <p align="center"><img alt="Switching a skill off for one project, then an MCP server from the Context tab" src="assets/switch.gif" width="900"></p>
- The Context tab asks Claude Code itself (`claude /context`) what fills the window at the start of a session in this
- folder, and lists the biggest consumers you control. Select one and press `x` to switch it off **for this project
- only**; the MCP and Skills tabs do the same and show each item's cost in tokens. catude uses the settings Claude Code
- documents for this (`disabledMcpServers`, `disabledMcpjsonServers`, `skillOverrides`): nothing is uninstalled, and `x`
- again turns it back on. For skills, `x` cycles through *name only*, *only when you type /name* and *off*.
- ## The agent map
- <p align="center"><img alt="Agents tab: every subagent with its runs, who launched it, tools, models and cost" src="assets/agents.gif" width="900"></p>
- Every subagent Claude Code can use here (yours, the project's, plugins' and the built-in ones) joined with what it
- really did in the last 30 days: how often it ran, who launched it (your conversation or another agent), the tools it
- used, the models, the tokens and an estimated cost. Agents you defined but never use are flagged, because their
- descriptions load into every session. `catude agents` prints the same map.
- ## Your own evidence when something feels worse
- <p align="center"><img alt="catude health: tool calls, failures, API retries, compactions, cache hit and tokens per response by day and version" src="assets/screenshots/cli-health.png" width="900"></p>
- `catude health` shows, day by day and per Claude Code version: tool calls and how many failed, API errors Claude
- Code retried, compactions, the share of input read from the cache, and fresh tokens per response. When a session
- feels worse, this is what changed on your machine.
- ## Audit what you installed
- <p align="center"><img alt="Audit tab: findings with the fix for each" src="assets/screenshots/audit.png" width="900"></p>
- Skills, hooks, plugins and MCP servers run with your permissions. 25 rules look for what usually goes wrong:
- passwords or tokens saved inside permission rules (what "don't ask again" can leave in `settings.local.json`),
- commands that delete data or rewrite history allowed without asking, plaintext secrets in MCP configs,
- `curl | sh`, prompt-injection phrases, hidden Unicode, unpinned `npx`, an `AGENTS.md` that Claude Code silently
- ignores. Every finding says why and what to do; `catude audit` exits with code 1 on high findings, so it fits in CI.
- [Rules](docs/security.md#audit-rules)
- ## Live: what every session is doing
- <p align="center"><img alt="Live tab: running sessions with their current tool, subagents, to-do list and context" src="assets/live.gif" width="900"></p>
- Each Claude Code session running on the machine (terminal, VS Code or Claude Desktop) as a card updated every
- second: working or waiting for you, the tool running and for how long, its last steps, what its subagents are doing,
- its to-do list, the model and effort, and how full the context is. In any tab, catude tells you when a session
- finishes and waits for you; turn on the bell or a system notification in preferences. [How it works](docs/live.md)
- ## Work profiles: one setup per kind of work
- <p align="center"><img alt="Profiles tab: frontend, backend and writing work profiles" src="assets/screenshots/profiles.png" width="900"></p>
- A work profile is a set of skills, agents, commands, MCP servers and plugins for one kind of work: frontend,
- backend, writing. Press `n` to build one, `a` to save your current setup, `s` to make it the one plain `claude`
- starts with. Start from one of the [15 presets below](#presets-and-where-they-come-from), load one from a file or
- URL (validated, pinned, hash-checked and audited first), or copy a profile's MCP servers into Claude Desktop with
- `catude profile desktop NAME`. [Guide](docs/profiles.md)
- ## Presets, and where they come from
- <p align="center"><img alt="Presets gallery: work profiles based on the public setups of known practitioners" src="assets/screenshots/presets.png" width="900"></p>
- Two kinds of presets, and every one says where it comes from. Nothing is invented: each preset quotes or links the
- public source it is based on, and catude shows that source before you apply anything.
- **Work profiles based on what practitioners published** (Profiles tab, `p`; `catude presets`). Only files under an
- open licence and pinned to a commit are installed; ideas without published files become a profile you complete.
- <table>
- <tr>
- <td width="33%" valign="top"><pre> ███ ████
- █ █ █ █
- █████ ████
- █ █ █ █
- █ █ █ █</pre><b>Armin Ronacher</b><br><sub>Simple, observable tooling (Makefiles, logs, CLIs) over clever automation</sub><br><br><a href="https://github.com/mitsuhiko/agent-stuff">mitsuhiko/agent-stuff</a><br><a href="https://lucumr.pocoo.org/2025/6/12/agentic-coding/">lucumr.pocoo.org</a><br><br><code>catude profile load preset:armin-ronacher</code></td>
- <td width="33%" valign="top"><pre>████ ████
- █ █ █
- ████ █
- █ █ █
- ████ ████</pre><b>Boris Cherny</b><br><sub>A mostly vanilla setup run in parallel; give Claude a way to verify its own work</sub><br><br><a href="https://threadreaderapp.com/thread/2007179832300581177">his threads on how the Claude Code team works</a><br><br><code>catude profile load preset:boris-cherny-claude-code</code></td>
- <td width="33%" valign="top"><pre>████ █ █
- █ █ █ █
- █ █ █████
- █ █ █ █
- ████ █ █</pre><b>Dex Horthy</b><br><sub>The whole workflow built around managing context: research, plan, implement</sub><br><br><a href="https://github.com/humanlayer/advanced-context-engineering-for-coding-agents/blob/main/ace-fca.md">humanlayer/advanced-context-engineering-for-coding-agents</a><br><a href="https://github.com/humanlayer/12-factor-agents">humanlayer/12-factor-agents</a><br><br><code>catude profile load preset:dex-horthy</code></td>
- </tr>
- <tr>
- <td width="33%" valign="top"><pre> ████ █ █
- █ █ █
- █ ██ █████
- █ █ █ █
- ████ █ █</pre><b>Geoffrey Huntley</b><br><sub>The "Ralph" loop: the same prompt fed to a fresh agent until the work is done</sub><br><br><a href="https://ghuntley.com/ralph/">ghuntley.com/ralph</a><br><a href="https://github.com/ghuntley/how-to-build-a-coding-agent">ghuntley/how-to-build-a-coding-agent</a><br><br><code>catude profile load preset:geoffrey-huntley-ralph</code></td>
- <td width="33%" valign="top"><pre>█ █ ████
- █ █ █ █
- █████ ████
- █ █ █ █
- █ █ █ █</pre><b>Harper Reed</b><br><sub>Spec, then plan, then small test-driven steps</sub><br><br><a href="https://github.com/harperreed/dotfiles/tree/master/.claude">harperreed/dotfiles</a><br><a href="https://harper.blog/2025/02/16/my-llm-codegen-workflow-atm/">harper.blog</a><br><br><code>catude profile load preset:harper-reed</code></td>
- <td width="33%" valign="top"><pre>█████
- █
- █
- █
- █████</pre><b>IndyDevDan</b><br><sub>Reusable prompts, subagents and hooks instead of vibe coding</sub><br><br><a href="https://github.com/disler/claude-code-hooks-mastery">disler/claude-code-hooks-mastery</a><br><br><code>catude profile load preset:indydevdan</code></td>
- </tr>
- <tr>
- <td width="33%" valign="top"><pre> ███ █ █
- █ █ █
- █ █ █
- █ █ █ █
- ███ █ </pre><b>Jesse Vincent</b><br><sub>Superpowers: mandatory skills for brainstorming, planning and TDD</sub><br><br><a href="https://github.com/obra/superpowers">obra/superpowers</a><br><a href="https://blog.fsck.com/2025/10/09/superpowers/">blog.fsck.com</a><br><br><code>catude profile load preset:jesse-vincent-superpowers</code></td>
- <td width="33%" valign="top"><pre>█ █ █ █
- █ █ █ █
- ███ ███
- █ █ █ █
- █ █ █ █</pre><b>Kieran Klaassen</b><br><sub>Compound engineering: every task makes the next one easier</sub><br><br><a href="https://github.com/EveryInc/compound-engineering-plugin">EveryInc/compound-engineering-plugin</a><br><br><code>catude profile load preset:kieran-klaassen-compound-engineering</code></td>
- <td width="33%" valign="top"><pre>█ █████
- █ █
- █ █
- █ █
- █████ █ </pre><b>Lauren Tan</b><br><sub>pstack: make one agent trustworthy on a whole problem before going wide</sub><br><br><a href="https://github.com/cursor/plugins/tree/main/pstack">cursor/plugins: pstack</a><br><br><code>catude profile load preset:lauren-tan-pstack</code></td>
- </tr>
- <tr>
- <td width="33%" valign="top"><pre>█ █ ████
- ██ ██ █ █
- █ █ █ ████
- █ █ █
- █ █ █ </pre><b>Matt Pocock</b><br><sub>Small, composable skills that keep you in control</sub><br><br><a href="https://github.com/mattpocock/skills">mattpocock/skills</a><br><br><code>catude profile load preset:matt-pocock-skills</code></td>
- <td width="33%" valign="top"><pre>█ █ █ █
- ██ ██ █ █
- █ █ █ █████
- █ █ █ █
- █ █ █ █</pre><b>Mitchell Hashimoto</b><br><sub>Real agents, not chat; when an agent fails, fix the harness</sub><br><br><a href="https://github.com/ghostty-org/ghostty/blob/main/AGENTS.md">ghostty AGENTS.md</a><br><a href="https://mitchellh.com/writing/my-ai-adoption-journey">mitchellh.com</a><br><br><code>catude profile load preset:mitchell-hashimoto</code></td>
- <td width="33%" valign="top"><pre>████ ████
- █ █ █
- ████ ███
- █ █
- █ ████ </pre><b>Peter Steinberger</b><br><sub>Short prompts, screenshots and many agents in parallel</sub><br><br><a href="https://github.com/steipete/agent-scripts">steipete/agent-scripts</a><br><a href="https://steipete.me/posts/just-talk-to-it">steipete.me</a><br><br><code>catude profile load preset:peter-steinberger</code></td>
- </tr>
- <tr>
- <td width="33%" valign="top"><pre> ████ █ █
- █ █ █
- ███ █ █ █
- █ ██ ██
- ████ █ █</pre><b>Simon Willison</b><br><sub>Agentic engineering patterns: spend the time saved on testing and review</sub><br><br><a href="https://simonwillison.net/guides/agentic-engineering-patterns/">simonwillison.net</a><br><br><code>catude profile load preset:simon-willison-agentic-engineering</code></td>
- <td width="33%" valign="top"><pre> ████ █ █
- █ █ █
- ███ █
- █ █
- ████ █ </pre><b>Steve Yegge</b><br><sub>Beads: a dependency-aware issue graph as the agent's memory</sub><br><br><a href="https://github.com/gastownhall/beads">gastownhall/beads</a><br><br><code>catude profile load preset:steve-yegge</code></td>
- <td width="33%" valign="top"><pre>█████ ████
- █ █ █
- █ ████
- █ █ █
- █ ████ </pre><b>Thorsten Ball</b><br><sub>An agent is an LLM, a loop and tools: brief it well</sub><br><br><a href="https://ampcode.com/how-to-build-an-agent">ampcode.com</a><br><br><code>catude profile load preset:thorsten-ball</code></td>
- </tr>
- </table>
- **30 configuration presets** (Config tab, `p`). Each one is a small group of documented settings for one goal, with
- the exact sentence of its source quoted next to it, and a preview of what it would change in your file:
- | Goal | Presets | Sources |
- |---|---|---|
- | Privacy | no telemetry or feedback uploads, block non-essential traffic, keep transcripts 7 days | [data usage](https://code.claude.com/docs/en/data-usage), [settings reference](https://code.claude.com/docs/en/settings-reference) |
- | Security | protect `.env` and credentials, strip credentials from subprocesses, no bypass mode, manual mode, audit setting changes | [Trail of Bits claude-code-config](https://github.com/trailofbits/claude-code-config), [permission modes](https://code.claude.com/docs/en/permission-modes) |
- | Sandbox | hide AWS/SSH credentials from commands, sandboxed Bash with fewer prompts, strict sandbox | [sandboxing](https://code.claude.com/docs/en/sandboxing) |
- | Lockdown | the enterprise-like strict example, and the lax one for long unattended runs | [anthropics/claude-code example settings](https://github.com/anthropics/claude-code/tree/main/examples/settings) |
- | Cost | Sonnet as default, subagents on Haiku, no background requests while idle | [costs](https://code.claude.com/docs/en/costs), [model config](https://code.claude.com/docs/en/model-config), [subagents](https://code.claude.com/docs/en/sub-agents) |
- | Git, UI, notifications, updates, Windows | ask before `git push`, no attribution, quieter UI, a bell or desktop notification when Claude needs you, stable channel, PowerShell tool | [hooks guide](https://code.claude.com/docs/en/hooks-guide), [setup](https://code.claude.com/docs/en/setup), [env vars](https://code.claude.com/docs/en/env-vars) |
- | Telemetry | your own usage metrics on a local Prometheus endpoint | [monitoring usage](https://code.claude.com/docs/en/monitoring-usage) |
- The **Discover** catalog follows the same rule: every plugin, skill and MCP server links to its repository and
- licence, and installs only through the official `claude plugin` and `claude mcp` commands.
- Presets reflect what each person or organisation published, at the link shown; they are not affiliated with or
- endorsed by them. A correction or a new source is a welcome [pull request](CONTRIBUTING.md).
- ## Every setting, and which one wins
- <p align="center"><img alt="Effective settings: the value Claude Code uses and the file it comes from" src="assets/screenshots/config-effective.png" width="900"></p>
- The Config tab lists every documented setting of every Claude app found (more than 300 settings and 400 environment
- variables), set or not, with what it does, its allowed values and its default. **Effective settings** shows, for each
- key of this project, the value Claude Code really uses, the file it comes from and what it overrides. Toggle, cycle
- or type a value; catude backs the file up, keeps its key order and refuses values that look like secrets. Press `p`
- for 30 [presets](docs/config.md#presets) (no telemetry, protect `.env`, lower cost, sandboxed Bash…), each quoting
- its source. [Details](docs/config.md)
- ## And also
- <table>
- <tr>
- <td width="50%"><img alt="Sessions" src="assets/screenshots/sessions.png"><br><b>Sessions</b>: every conversation of the terminal and of Claude Desktop's Code tab, with its exact context, an estimated cost and a summary without AI (your first message, files changed, tools). <code>n</code> names it, <code>c</code> continues it in Claude Code, <code>f</code> searches every conversation.</td>
- <td width="50%"><img alt="Models" src="assets/screenshots/models.png"><br><b>Models</b>: which Claude model and effort level is stronger for a kind of work, on public benchmarks, with cost per task and advice on the one you use. No account or key.</td>
- </tr>
- <tr>
- <td width="50%"><img alt="Graph" src="assets/screenshots/graph.png"><br><b>Graph</b>: apps, config files, instructions, skills, hooks, MCP servers, the programs they start and running processes, as a tree; <code>w</code> opens it as a diagram.</td>
- <td width="50%"><img alt="Discover" src="assets/screenshots/discover-tag.png"><br><b>Discover</b>: 177 plugins, 89 skills, 55 marketplaces and 25 MCP servers, checked at their source and tagged; installed with the official <code>claude plugin</code> command.</td>
- </tr>
- </table>
- | | |
- |---|---|
- | **Instructions** | `catude instructions` says which `CLAUDE.md`, `CLAUDE.local.md` and `AGENTS.md` Claude Code reads here, and the project's auto memory. A `CLAUDE.local.md` silently stops `AGENTS.md`: `--fix` adds the import, and copies `.agents/skills` (which Claude Code never reads) into `.claude/skills` |
- | **Every extension, explained** | Skills, agents, commands, hooks, MCP servers and plugins from every scope, each tab with one sentence on what they are; `e` edits a file after a backup |
- | **Storage** | Transcripts, caches and Claude Desktop data with their sizes, safe cleaning, and a warning before Claude Code deletes old transcripts |
- | **Export and import** | Your setup in one file without secrets, to move it or share it; importing shows what it brings and audits it first ([format](docs/rigcard.md)) |
- | **Inside Claude Code** | The repository is a plugin marketplace: `/catude:doctor`, `/catude:audit`, `/catude:sessions`, `/catude:models` |
- | **10 languages** | English, Español, 日本語, 简体中文, 繁體中文, 한국어, Deutsch, Français, Português (Brasil), Русский, with the terms of Anthropic's localized docs ([how](docs/translations.md)) |
- | **Any terminal** | Detects what your terminal and font can draw, themes including high contrast, reduced motion, a lite view (`ctlite`) ([accessibility](docs/accessibility.md)) |
- <details>
- <summary><b>A walk through every tab</b></summary>
- <p align="center"><img alt="A walk through all the tabs" src="assets/tour.gif" width="900"></p>
- | | |
- |---|---|
- |  Overview |  Config |
- |  Live |  Context |
- |  Usage |  Models |
- |  Profiles |  Sessions |
- |  Graph |  Skills |
- |  Agents |  Commands |
- |  MCP |  Plugins |
- |  Hooks |  Discover |
- |  Tools |  Storage |
- |  Audit |  Preferences |
- |  `catude agents` |  Help (`?`) |
- </details>
- ## Command line
- Everything in the app is also a command, and every listing accepts `--json`.
- ```sh
- catude why # where your tokens went in the last 14 days, and what to change
- catude health # tool calls, failures, API retries, compactions, cache hit, by day and version
- catude agents # the agent map: every subagent, its runs, who launched it, tokens
- catude instructions [--fix] # which CLAUDE.md, CLAUDE.local.md and AGENTS.md Claude Code reads here
- catude audit # security findings (exit code 1 if any is high)
- catude context # what fills your context window, measured by Claude Code
- catude doctor # every file Claude reads here, and whether it exists
- catude ls mcp --live # MCP servers including claude.ai connectors, with health
- catude profile new web --skills frontend-design,tdd --mcp playwright
- ct switch # pick the active work profile from a list
- eval "$(catude shell-init zsh)" # from now on, plain `claude` starts the active profile
- catude models compare "opus 5.5 high" "fable 5.2 medium" --for coding
- catude export my-setup.json # your setup in one file, without secrets
- ```
- Inside Claude Code: `/plugin install catude --marketplace catude/catude` adds `/catude:doctor`,
- `/catude:audit`, `/catude:sessions` and `/catude:models`. Keys and every command: [docs/usage.md](docs/usage.md).
- ## Privacy and safety
- 1. **Local and read-only by default.** catude reads files on your machine. Every write is explicit, backed up
- first and reversible.
- 2. **No telemetry, account or API key.** It goes online only when you ask for something that needs it: importing
- a URL you name, installing through the official `claude` commands, or the public model data of the Models tab.
- 3. **Secrets stay masked.** Values that look like credentials are shown as `••••` and never exported. The session
- key files of Claude Code are never opened.
- 4. **Official commands and documented settings.** Plugins and MCP servers are installed with `claude plugin` and
- `claude mcp`; per-project switches use the settings Anthropic documents.
- 5. **Measured numbers.** When a number is an estimate (a cost at list prices), catude says so.
- ### Security you can check
- | | |
- |---|---|
- | **Every change is scanned** | [CodeQL](https://github.com/catude/catude/actions/workflows/codeql.yml) (code scanning), `govulncheck` (known vulnerabilities in Go and its dependencies) and `gitleaks` (secrets) run on every pull request and push, next to the tests on Windows, macOS and Linux ([ci.yml](.github/workflows/ci.yml)) |
- | **Protected main branch** | Changes arrive through pull requests that must pass every check; no force pushes |
- | **Dependencies kept current** | Dependabot opens updates for Go modules and GitHub Actions; releases are built with the latest Go and its security fixes |
- | **Verifiable downloads** | Every release has a `checksums.txt` and a signed build provenance attestation: `gh attestation verify catude_*.zip --repo catude/catude` proves a file was built here, from this source |
- | **Small surface** | One static binary, no runtime, no telemetry, no account; it goes online only for what you ask |
- [Security model](docs/security.md) · [Report a vulnerability](SECURITY.md) (privately, through GitHub)
- ## FAQ
- <details>
- <summary><b>Are the token numbers the same as my plan limits?</b></summary>
- No. Plan limits are computed by Anthropic and also count your other devices and claude.ai. catude shows what this
- machine recorded, which is what you can act on: the same responses, de-duplicated, with where they went.
- </details>
- <details>
- <summary><b>Does it change my Claude configuration?</b></summary>
- Only when you press a key that does so (switch something off, change a setting, install a profile, clean a folder).
- Before each change it copies the file to `<user cache>/catude/backups/` and it leaves the rest of the file untouched.
- </details>
- <details>
- <summary><b>Is it an official Anthropic tool?</b></summary>
- No. catude is an independent open-source project, not affiliated with or endorsed by Anthropic. The paths and
- settings it reads come from Anthropic's documentation.
- </details>
- <details>
- <summary><b>I use the Microsoft Store version of Claude Desktop. Does it work?</b></summary>
- Yes. The Store app reads a virtualised copy of its configuration under `%LOCALAPPDATA%\Packages\Claude_*\…`, so
- editing `%APPDATA%\Claude\` has no effect. catude shows and edits the copy the app actually reads.
- </details>
- <details>
- <summary><b>Where do the screenshots come from?</b></summary>
- From an invented machine (user `alex`, invented projects, sessions and usage) built by
- [scripts/demo](scripts/demo/README.md). The command-line pictures are the real output of catude on that machine.
- </details>
- ## Alternatives
- Other tools cover parts of this, among them [ccusage](https://github.com/ccusage/ccusage) (usage reports),
- [cc-switch](https://github.com/farion1231/cc-switch) (providers), [claude-code-templates](https://github.com/davila7/claude-code-templates)
- (a catalog) and [claude-swap](https://github.com/realiti4/claude-swap) (accounts, which catude uses for plan limits).
- [docs/comparison.md](docs/comparison.md) says what each does better.
- ## Contributing
- Pull requests are very welcome, and these need no Go: a [configuration preset](internal/claudecfg/config-presets.json)
- with its public source, a [catalog entry](internal/catalog/catalog.json), a [practitioner preset](internal/presets/presets.json),
- a translation fix backed by a source in the [glossaries](docs/translations.md), or a bug report with the output of
- `catude debug` (it contains no secrets). Code: [CONTRIBUTING.md](CONTRIBUTING.md). Roadmap: [docs/roadmap.md](docs/roadmap.md).
- ---
- <sub>catude is an independent project, not affiliated with or endorsed by Anthropic. "Claude" is a trademark of
- Anthropic, PBC. Benchmark data © Epoch AI (CC BY 4.0) and the projects it credits. MIT © catude contributors.</sub>
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.