REA MCP config generator

Unofficial guide. This site is an independent community guide and is not affiliated with the REA project. For authoritative docs see the official REA website and the morluto/rea GitHub repository. Commands here are taken from the official README/docs (checked 2026-10-19); REA changes quickly, so verify against the official docs.

Pick your agent to get the exact config block that rea setup writes, the file it goes in, and the one-line command that does it for you. Every snippet was captured from a real rea-agents 6.5.0 setup run on 2026-10-17. Nothing is hand-written.

Agent rea-agents version

1. Easiest: let setup write it

2. Or edit the file yourself

File: (Linux path from our test run. On macOS and Windows some apps, such as Claude Desktop and VS Code, keep this file elsewhere; setup --dry-run prints the right path for your machine.)

How this works

REA is an MCP server. Each agent keeps its list of MCP servers in its own config file, in its own format. rea setup knows the location and format for every supported client. To capture the snippets above, we ran setup in an empty test home folder with all 24 supported --client values and copied the files it created. That's why some entries differ. VS Code uses a top-level servers key with "type": "stdio". OpenCode puts the whole command in one array under mcp. GitHub Copilot CLI adds "tools": ["*"]. Codex and Grok Build use TOML with startup_timeout_sec = 30. Kimi Code sets startupTimeoutMs and toolTimeoutMs. Hermes uses YAML.

Every client runs the same server command: npx -y rea-agents@<version> mcp. Setup pins an exact version so the tool list doesn't change between sessions. When a new release comes out, run setup again (or change the version above) to move the pin.

Setup command or manual edit?

Use the setup command when you can. According to the official installation docs, setup keeps unrelated entries and comments, makes a .rea.backup before editing, replaces the file atomically and checks the result. It also installs REA's workflow skill (skip it with --skill=false). It refuses to touch a malformed config instead of guessing. Edit the file by hand only if your agent's config lives somewhere non-standard or is managed by another tool. If the file already has other servers, add the rea entry next to them; don't replace the whole file.

Two things we saw in testing. First, setup run from a script without --yes stops at status: needs_confirmation and writes nothing. Second, for Claude Desktop the plan warns: "Quit Claude Desktop fully before approving; closing its window is not enough." Otherwise the running app can rewrite the file and drop the entry.

After you add it

Restart the agent.
Run npx -y [email protected] doctor --client <id>. Look for healthy: true and a registration:<id> check with ok: true.
If your model complains about too many tools or context size, add REA_MCP_INPUT_SCHEMA_PROFILE=compact to the server's environment (details). On Gemini models, also read the 400 INVALID_ARGUMENT page.

Your Node.js must be 22.19+, 24.11+ or 26+ (install guide).

FAQ

Where is the REA config for Claude Code?

~/.claude.json, under mcpServers.rea, with command npx and args -y rea-agents@VERSION mcp.

Where is the REA config for Cursor?

~/.cursor/mcp.json, under mcpServers.rea.

Where is the REA config for Codex?

~/.codex/config.toml as [mcp_servers.rea] with startup_timeout_sec = 30.

Why does VS Code look different?

VS Code's mcp.json uses a top-level servers key and type stdio instead of mcpServers.