Skip to content
Back to search

concord-mcp

Live messaging for coding agents

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/Get-Concord-AI/concord-mcp 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/Get-Concord-AI/concord-mcp'

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/get-concord-ai--concord-mcp/badge.svg)](/listings/get-concord-ai--concord-mcp)

HTML

<a href="/listings/get-concord-ai--concord-mcp"><img src="/api/listings/get-concord-ai--concord-mcp/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.

npm install -g @concord-ai/concord-mcp
cd /path/to/your/repository
concord setup

Restart your agent clients, then ask two of them to work in the same repository. Concord gives them a shared workspace and makes reachable sessions available for direct prompts and replies.

<details> <summary>What <code>concord setup</code> changes</summary>

`concord setup` creates the local `.concord/` workspace, registers the MCP server for Claude, Cursor, Gemini, Grok, and Codex (`.mcp.json`, `.cursor/mcp.json`, `.gemini/settings.json`, `.grok/config.toml`, and `~/.codex/config.toml`) and writes Concord's tool instructions into your client configs (`CLAUDE.md`, `AGENTS.md`, `.codex/`, `.cursor/rules/`). It merges into existing config rather than replacing it and is safe to re-run.

Setup also detects supported clients and attempts to install their global Concord adapters independently. Use `--no-adapters` to skip that step or `--require-adapters` in managed installs that should fail on degraded support. Pass `--no-mcp` to write only the workspace and instructions while managing MCP registration yourself.

</details>

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 · 366 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. <p align="center">
  2. <a href="https://getconcord.ai">
  3. <img src="./assets/concord-readme-header.png" alt="Concord MCP — shared work-state for coding agents" width="100%">
  4. </a>
  5. </p>
  6. <h1 align="center">Concord MCP</h1>
  7. <p align="center"><strong>Let Claude Code, Codex, Cursor, Gemini CLI, Grok Build, Goose, GitHub Copilot CLI, and OpenCode talk to each other.</strong></p>
  8. <p align="center">
  9. The open-source, local-first communication and coordination layer for AI coding agents.
  10. Send live messages across harnesses, detect overlapping work before agents edit,
  11. share decisions, and hand off tasks with evidence - through one MCP server.
  12. </p>
  13. <p align="center">
  14. <a href="https://www.npmjs.com/package/@concord-ai/concord-mcp"><img src="https://img.shields.io/npm/v/@concord-ai/concord-mcp.svg" alt="npm version"></a>
  15. <a href="https://www.npmjs.com/package/@concord-ai/concord-mcp"><img src="https://img.shields.io/npm/dm/@concord-ai/concord-mcp.svg" alt="npm downloads"></a>
  16. <a href="https://github.com/Get-Concord-AI/concord-mcp/actions/workflows/ci.yml"><img src="https://github.com/Get-Concord-AI/concord-mcp/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
  17. <a href="https://www.npmjs.com/package/@concord-ai/concord-mcp"><img src="https://img.shields.io/node/v/@concord-ai/concord-mcp.svg" alt="Node.js version"></a>
  18. <a href="./LICENSE"><img src="https://img.shields.io/badge/License-MIT-blue.svg" alt="MIT License"></a>
  19. </p>
  20. <p align="center">
  21. <a href="#see-it-work">Demo</a> ·
  22. <a href="#quick-start">Quick start</a> ·
  23. <a href="#supported-agents">Supported agents</a> ·
  24. <a href="https://getconcord.ai">Website</a> ·
  25. <a href="./CONTRIBUTING.md">Contributing</a>
  26. </p>
  27. <p align="center">
  28. <a href="./README.md">English</a> |
  29. <a href="./README.zh-CN.md">简体中文</a> |
  30. <a href="./README.zh-TW.md">繁體中文</a> |
  31. <a href="./README.ja.md">日本語</a> |
  32. <a href="./README.ko.md">한국어</a> |
  33. <a href="./README.es.md">Español</a> |
  34. <a href="./README.pt-BR.md">Português</a> |
  35. <a href="./README.fr.md">Français</a> |
  36. <a href="./README.de.md">Deutsch</a> |
  37. <a href="./README.ru.md">Русский</a>
  38. </p>
  39. ## See it work
  40. Two agents in different harnesses can discover each other, exchange messages,
  41. and divide work without a human relaying context between them:
  42. ```text
  43. Claude Code → Concord Claim src/app/page.tsx
  44. Codex → Concord Claim src/app/page.tsx
  45. Concord → Codex Overlap: Claude Code already owns this file
  46. Codex → Claude I'll take src/app/api instead. Does that work?
  47. Claude → Codex Yes. I'll keep the page and use your API contract.
  48. ```
  49. Run the [real Claude Code ↔ Codex demo](./examples/whack-a-mole/) to watch both
  50. agents resolve an overlapping claim through a live prompt/reply, build a playable
  51. app, transfer ownership, and hand the result to an independent reviewer.
  52. ## Quick start
  53. ```bash
  54. npm install -g @concord-ai/concord-mcp
  55. cd /path/to/your/repository
  56. concord setup
  57. ```
  58. No Node.js? On macOS or Linux, install a self-contained build instead:
  59. ```bash
  60. curl -fsSL https://getconcord.ai/install.sh | sh
  61. ```
  62. It bundles its own Node runtime under `~/.concord/standalone` and puts the
  63. `concord` and `concord-mcp` launchers in `~/.local/bin` (override with
  64. `CONCORD_INSTALL_DIR` / `CONCORD_BIN_DIR`; pin a release with `CONCORD_VERSION`).
  65. Restart your agent clients, then ask two of them to work in the same repository.
  66. Concord gives them a shared workspace and makes reachable sessions available for
  67. direct prompts and replies.
  68. <details>
  69. <summary>What <code>concord setup</code> changes</summary>
  70. `concord setup` creates the local `.concord/` workspace, registers the MCP server
  71. for Claude, Cursor, Gemini, Grok, and Codex (`.mcp.json`, `.cursor/mcp.json`,
  72. `.gemini/settings.json`, `.grok/config.toml`, and `~/.codex/config.toml`).
  73. For [Goose](./docs/goose.md), it registers the `concord-relay` extension in
  74. Goose's `config.yaml` when detected. For
  75. [Antigravity](./docs/antigravity.md), it writes `.agents/mcp_config.json`,
  76. `.agents/hooks.json`, and `.agents/rules/concord.md`. It also writes
  77. Concord's tool instructions into your client configs (`CLAUDE.md`, `AGENTS.md`,
  78. `.codex/`, `.cursor/rules/`). It merges into existing config rather than replacing
  79. it and is safe to re-run.
  80. Setup also detects supported clients and attempts to install their global Concord
  81. adapters independently. Use `--no-adapters` to skip that step or
  82. `--require-adapters` in managed installs that should fail on degraded support.
  83. Pass `--no-mcp` to write only the workspace and instructions while managing MCP
  84. registration yourself.
  85. With global mode on (see below), setup leaves the repository's instruction and
  86. MCP files alone, because the user-level registrations already cover them; pass
  87. `--repo-files` to write them anyway, for example to check them in for a team.
  88. </details>
  89. <details>
  90. <summary>Using Concord in repositories you have not set up</summary>
  91. The first time you run `concord setup` interactively, it asks whether your agents
  92. should use Concord in repositories where setup was never run (default yes). Your
  93. answer is saved in `~/.concord/config.json`; `concord setup --global` or
  94. `--no-global` changes it.
  95. With global mode on, Concord registers itself at user level for each installed
  96. agent and adds a short Concord section to their user instruction files
  97. (`~/.claude/CLAUDE.md`, `~/.codex/AGENTS.md`, `~/.gemini/GEMINI.md`, Goose's
  98. `.goosehints`, `~/.copilot/copilot-instructions.md`; Cursor receives it from its global session hook). Nothing is
  99. written into those repositories: their work is kept in `~/.concord/global.db`,
  100. one project per repository, named by its git remote as Concord Cloud names it.
  101. Sessions outside any repository share the `general` project. A repository you
  102. did set up keeps everything in its own `.concord/`; `concord setup` there writes
  103. no instruction or MCP files while global mode is on unless you pass
  104. `--repo-files`. Turning global mode off removes exactly what it added.
  105. </details>
  106. <details>
  107. <summary>Concord Cloud: agents on different machines</summary>
  108. `concord cloud login` signs this machine in to Concord Cloud and links it. A
  109. linked machine works through the cloud in every folder, so its agents reach
  110. agents on any machine. No repository keeps local state while the machine is
  111. linked; a repository with a git remote is that remote's project, one without a
  112. remote is a project of its own on this machine, and anything outside a
  113. repository is `general`.
  114. `concord cloud unlink` switches the whole machine back to local Concord while
  115. keeping the login, and `concord cloud link` switches it back on.
  116. `concord cloud status` says which applies, and `concord doctor` checks the link,
  117. and lists local state the machine no longer uses, with the commands to move it
  118. aside once no session started before the link is running (nothing is deleted).
  119. </details>
  120. ## Supported agents
  121. | Agent | Integration guide |
  122. | ---------------------------------- | -------------------------------------------------- |
  123. | Claude Code | [Setup and delivery](./docs/claude-code.md) |
  124. | Codex | [Setup and delivery](./docs/codex.md) |
  125. | Cursor | [Setup and delivery](./docs/cursor.md) |
  126. | Gemini CLI | [Setup and delivery](./docs/gemini-cli.md) |
  127. | Grok Build | [Setup and delivery](./docs/grok-build.md) |
  128. | Goose | [Setup and delivery](./docs/goose.md) |
  129. | Google Antigravity | [Setup and delivery](./docs/antigravity.md) |
  130. | GitHub Copilot CLI | [Setup and delivery](./docs/github-copilot-cli.md) |
  131. | OpenCode | [Setup and delivery](./docs/opencode.md) |
  132. | Devin CLI | Project-level setup via `concord setup` |
  133. | Conductor (Claude Code, Codex) | [Setup and delivery](./docs/conductor.md) |
  134. | Any other MCP-capable coding agent | [Manual setup](./docs/other-mcp-clients.md) |
  135. > There is no universal `/concord` slash command — commands are client-specific.
  136. > Concord works through MCP tools plus the installed instructions on any
  137. > MCP-capable client.
  138. Live delivery depends on the receiving harness and session state. Run
  139. `concord adapters status` to see which installed agents are reachable and how
  140. messages will be delivered.
  141. ## Communication is the starting point
  142. Messaging gets agents talking. Concord's shared work-state keeps the resulting
  143. collaboration reliable after the message is delivered.
  144. | Without Concord | With Concord |
  145. | --------------------------------------------------- | -------------------------------------------------------------- |
  146. | Agents cannot contact peers in another harness | Agents send direct, replyable prompts across supported clients |
  147. | Agents discover collisions after editing | Agents claim files and modules before work begins |
  148. | Context disappears when a session ends | Decisions, assumptions, and findings stay attached to the task |
  149. | Ownership is implied by chat history | Assignments and handoffs are explicit and acknowledged |
  150. | Humans reconstruct progress from branches and diffs | Review packets arrive with scope, tests, risks, and provenance |
  151. Concord is not another autonomous agent or orchestrator. It is the shared layer
  152. around your agents: presence, messaging, task memory, ownership, handoffs, and
  153. review state through one small MCP server.
  154. ## The tools
  155. | Tool | Purpose |
  156. | --------------- | ------------------------------------------------------------------------------------------- |
  157. | `start_work` | registers presence, claims or accepts one task, and reports scope overlaps before editing |
  158. | `inspect_work` | reads workspace/task state, an agent inbox/outbox, or a durable prompt/reply thread |
  159. | `update_work` | records task context or immediately prompts/replies to another promptable workspace agent |
  160. | `transfer_work` | assigns, accepts, declines, releases, reassigns, offers handoffs, or reopens versioned work |
  161. | `finish_work` | records evidence and optionally marks a task review-ready, complete, or closed |
  162. Writes accept an `agent_id`, which keeps presence live just by working.
  163. `inspect_work` shows **who is here** with three separate signals — heartbeat,
  164. last activity, and the status the agent reported (writing never changes the
  165. reported status) — and flags **stale claims**: work whose owning agent has gone
  166. away without handing off, an assignment nobody accepted, or in-progress work
  167. with no owning agent at all.
  168. For live agent-to-agent communication, run `concord setup`, then restart existing
  169. client sessions once. A prompt uses `update_work` with `operation: "prompt"`, the
  170. target `to_agent_id`, content, and an `idempotency_key`; a reply uses
  171. `operation: "reply"` and `reply_to_message_id`. A receipt-bearing adapter steers
  172. a busy turn or starts an idle turn. Hook-only integrations leave a durable pull
  173. message and state that limitation in the result. Delivery fails immediately
  174. when the named agent has no reachable endpoint; Concord does not silently
  175. reroute it.
  176. `concord adapters status` reports each harness separately, including its
  177. monitor/controller kind, verified reachability, required action, and version
  178. probe result. `concord adapters install`, `doctor`, and `uninstall` provide the
  179. same global lifecycle outside repository setup.
  180. Concord resolves the repository workspace automatically. Operations return its
  181. `workspace_id` and repository root so a client can detect a misrouted call; the
  182. id can be passed explicitly when one server is coordinating multiple roots.
  183. Lifecycle-changing operations use the task's monotonic `version` as
  184. `expected_version`. If two agents act on the same version, only the first
  185. transition succeeds. Assignment leaves work in `assigned` until the named agent
  186. uses `transfer_work` with `action: "accept"`; a handoff offer likewise keeps
  187. ownership with the sender until the recipient accepts. Every ownership change
  188. is retained in an append-only audit history.
  189. ## What you get
  190. SQLite is the local source of truth, kept in the `.concord/` at the **root of
  191. the repo** the work is happening in. The MCP server resolves that root from
  192. `CONCORD_REPO_ROOT` if set, then Claude Code's `CLAUDE_PROJECT_DIR` (which Claude
  193. Code sets automatically, even for a user-scoped server), then its working
  194. directory — so every agent in one repo shares one store. Set `CONCORD_REPO_ROOT`
  195. when running the server somewhere its working directory is not inside the repo.
  196. Linked Git worktrees follow Git's `commondir` metadata to the primary checkout,
  197. so the main checkout and all linked worktrees intentionally share one Concord
  198. database and workspace id.
  199. To restrict explicit workspace selection, set `CONCORD_ALLOWED_ROOTS` to a
  200. path-delimited list of allowed repository roots. Without an allowlist, decoded
  201. roots must still exist and be directories.
  202. `concord setup` writes a `.concord/.gitignore` containing `*`, so the generated
  203. workspace ignores itself and stays local without editing the repository's shared
  204. `.gitignore`. Teams that want selected artifacts in PRs can add negation rules to
  205. that file (setup keeps existing rules, only ensuring the `*` catch-all) or
  206. force-add the human-readable files. Repositories set up by Concord 0.10.5 or
  207. earlier also have a `.concord/` rule in the root `.gitignore`; remove it before
  208. negations in `.concord/.gitignore` can take effect.
  209. ```text
  210. .concord/
  211. ├── .gitignore ignores the whole directory
  212. ├── concord.db local source of truth
  213. ├── HANDOFF.md human-readable handoff
  214. ├── REVIEW_PACKET.md review-ready evidence
  215. ├── events.jsonl chronological tool events, one JSON object per line
  216. └── WORK_STATE.json generated work-state export
  217. ```
  218. ## CLI
  219. Concord supports both typed MCP tools and a regular CLI. MCP-capable agents can
  220. call the tools directly; humans and CLI-oriented agents can work with the same
  221. shared workspace through `concord` commands.
  222. ```bash
  223. concord setup # set up local state, instructions, and MCP clients
  224. concord status # roster, active work, overlaps, stale claims, review-ready
  225. concord dashboard # live, keyboard-driven view of agents, tasks, alerts, and activity
  226. concord who # which agents are present and what they are working on
  227. concord tasks # list all tracked tasks
  228. concord handoff <task-id> # print the latest handoff
  229. concord review-packet <id> # print the latest review packet
  230. concord export markdown # regenerate .concord/ artifacts
  231. concord check <files...> # check exact-file overlaps with active tasks (for hooks/CI)
  232. concord watch # print work-state whenever .concord changes (Ctrl-C to stop)
  233. concord doctor # workspace checks + per-task tool adoption
  234. concord adapters status # global harness delivery capability matrix
  235. concord --repo ../project status # select by repository path from anywhere
  236. concord --workspace ws_... status # select an id returned by a Concord operation
  237. ```
  238. To block conflicting edits in a pre-commit hook or CI, run `concord check src/file.ts`
  239. (or `concord check --task <id> src/file.ts` to exclude your own task).
  240. The command exits with status `1` when an active task claims the same file,
  241. and `0` when there is no overlap. Claude Code PreToolUse hooks treat exit code
  242. `2` as a denial, so use `concord check src/file.ts || exit 2` in that hook.
  243. `--repo` and `--workspace` are global, mutually exclusive options. The CLI uses
  244. the same `CONCORD_REPO_ROOT` → `CLAUDE_PROJECT_DIR` → working-directory priority
  245. and the same linked-worktree canonicalization as MCP.
  246. `concord dashboard` is a read-only, full-screen local TUI. It refreshes from the
  247. shared SQLite workspace every second while keeping agents, tasks, alerts,
  248. context, and timeline inside a fixed terminal viewport. Use `Tab` to change
  249. panes, `j`/`k` or the arrow keys to select work, `/` to filter, `?` for help,
  250. and `q` to quit.
  251. ## Upgrade
  252. ```bash
  253. npm install -g @concord-ai/concord-mcp@latest
  254. concord --version
  255. ```
  256. Rerun the `curl … | sh` installer to upgrade a self-contained install.
  257. Concord checks daily and surfaces available updates in the CLI, MCP tools, and
  258. dashboard; `concord setup` can install one with confirmation (with npm or the installer, whichever you used), and
  259. `CONCORD_NO_UPDATE_CHECK=1` disables checks.
  260. ## What this is / is not
  261. Shared work-state and task memory for coding agents using the same local
  262. checkout. **Not** an orchestrator, code reviewer, hosted sync service, memory
  263. vector DB, or autonomous coding agent.
  264. See also: [Why not just use markdown?](./docs/why-not-markdown.md)
  265. ## Contributing
  266. See [`CONTRIBUTING.md`](./CONTRIBUTING.md) and [`CLAUDE.md`](./CLAUDE.md). This
  267. repo is strictly typed (no `any`, no typecasts) and modular. Good first issues
  268. are labelled [`good first issue`](https://github.com/Get-Concord-AI/concord-mcp/labels/good%20first%20issue).
  269. ## Star History
  270. <a href="https://www.star-history.com/?repos=Get-Concord-AI%2Fconcord-mcp&type=date&legend=top-left">
  271. <picture>
  272. <source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/chart?repos=Get-Concord-AI/concord-mcp&type=date&theme=dark&legend=top-left&sealed_token=PLqZk9LDZSite53TJb5bwyYask4-B5wAFxLcp3Ki_mJ3KlMtSpH09lN3ohwJuoQNs5Kwg_0zPG2DEwcRzJZvCIP1Md6vMdHjNwF_WkjuX6gd44wkwWFzqrAWgXGx2bPFEuJFRhmuWUlbRLnPOw6uW7cVJzrlcxprXCDU3_hHmUlHaFjVjQulH7EkfyuA" />
  273. <source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/chart?repos=Get-Concord-AI/concord-mcp&type=date&legend=top-left&sealed_token=PLqZk9LDZSite53TJb5bwyYask4-B5wAFxLcp3Ki_mJ3KlMtSpH09lN3ohwJuoQNs5Kwg_0zPG2DEwcRzJZvCIP1Md6vMdHjNwF_WkjuX6gd44wkwWFzqrAWgXGx2bPFEuJFRhmuWUlbRLnPOw6uW7cVJzrlcxprXCDU3_hHmUlHaFjVjQulH7EkfyuA" />
  274. <img alt="Star History Chart" src="https://api.star-history.com/chart?repos=Get-Concord-AI/concord-mcp&type=date&legend=top-left&sealed_token=PLqZk9LDZSite53TJb5bwyYask4-B5wAFxLcp3Ki_mJ3KlMtSpH09lN3ohwJuoQNs5Kwg_0zPG2DEwcRzJZvCIP1Md6vMdHjNwF_WkjuX6gd44wkwWFzqrAWgXGx2bPFEuJFRhmuWUlbRLnPOw6uW7cVJzrlcxprXCDU3_hHmUlHaFjVjQulH7EkfyuA" />
  275. </picture>
  276. </a>
  277. ## Privacy & telemetry
  278. Concord sends product and coordination telemetry to `getconcord.ai`. It includes
  279. random installation/invocation identifiers; irreversible per-install workspace
  280. and task-flow pseudonyms; Concord/Node/platform versions; normalized client
  281. metadata; operation names, outcomes, and durations; aggregate overlap/edit-guard
  282. results; message delivery stages and latencies; task lifecycle transitions and
  283. elapsed time; and explicitly reported acceptance,
  284. integration, human-intervention, and rework outcomes.
  285. Concord never sends code, raw file or repository paths, remotes, usernames, raw
  286. task or agent identifiers, message identifiers or content, command arguments,
  287. tool inputs/outputs, or task content. The receiving server stores the request IP
  288. address and derives/stores a country code. Those server-side fields currently
  289. have no automatic expiry. Set `CONCORD_TELEMETRY_DISABLED=1` (or
  290. `DO_NOT_TRACK=1`) to disable telemetry. Delivery is best effort and can never
  291. make a Concord operation fail.
  292. ## License
  293. [MIT](./LICENSE)

Ratings and comments

Sign in to rate this skill.

Sign in

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