DOCS

CONNECT IN ONE PASTE.

Sign up, then tell Setup where your agent runs. In Claude Desktop or claude.ai, add Scout as a connector and sign in. In a terminal, create a key and paste one command. Then add one line to your agent's instruction file so it reads Scout first.

{ "mcpServers": { "scout": { "url": "https://api.runscout.dev/mcp/your-team",
 "headers": { "Authorization": "Bearer ${SCOUT_API_KEY}" } } } }

Claude Code

Terminal — also covers the Code tab in the Claude desktop app

claude mcp add --transport http --scope user scout https://api.runscout.dev/mcp/your-team --header "Authorization: Bearer $SCOUT_API_KEY"

Run it in a terminal where SCOUT_API_KEY is set. --scope user adds Scout for every project and keeps the key in your user settings, not in the repository. Restart any open Claude Code session afterwards.

Add one line to your agent's instruction file, so it reads Scout before anything else and goes with Scout when the two disagree. Models give that file more weight than anything a tool returns.

At the start of every session, call scout_session_bundle and read it before you plan or change anything. When this file or the code disagrees with Scout's current decision, follow Scout's and say so.

Where it goes: ~/.claude/CLAUDE.md for Claude Code, ~/.codex/AGENTS.md for Codex, and an always-applied rule in .cursor/rules for Cursor.

Full guide: Scout for Claude Code

Claude Desktop

Claude → Settings → Developer → Edit Config (claude_desktop_config.json)

{
  "mcpServers": {
    "scout": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.runscout.dev/mcp/your-team", "--header", "Authorization:${AUTH_HEADER}"],
      "env": { "AUTH_HEADER": "Bearer <YOUR_API_KEY>" }
    }
  }
}

Claude Desktop cannot read SCOUT_API_KEY from your shell, so replace <YOUR_API_KEY> with your key in this file. Needs Node.js installed. Quit Claude fully and reopen it afterwards.

Add one line to your agent's instruction file, so it reads Scout before anything else and goes with Scout when the two disagree. Models give that file more weight than anything a tool returns.

At the start of every session, call scout_session_bundle and read it before you plan or change anything. When this file or the code disagrees with Scout's current decision, follow Scout's and say so.

Where it goes: ~/.claude/CLAUDE.md for Claude Code, ~/.codex/AGENTS.md for Codex, and an always-applied rule in .cursor/rules for Cursor.

Cursor

~/.cursor/mcp.json, or .cursor/mcp.json in one project

{
  "mcpServers": {
    "scout": {
      "url": "https://api.runscout.dev/mcp/your-team",
      "headers": { "Authorization": "Bearer ${env:SCOUT_API_KEY}" }
    }
  }
}

Add one line to your agent's instruction file, so it reads Scout before anything else and goes with Scout when the two disagree. Models give that file more weight than anything a tool returns.

At the start of every session, call scout_session_bundle and read it before you plan or change anything. When this file or the code disagrees with Scout's current decision, follow Scout's and say so.

Where it goes: ~/.claude/CLAUDE.md for Claude Code, ~/.codex/AGENTS.md for Codex, and an always-applied rule in .cursor/rules for Cursor.

Full guide: Scout for Cursor

Codex

Terminal, or ~/.codex/config.toml

In a terminal

codex mcp add scout --url https://api.runscout.dev/mcp/your-team --bearer-token-env-var SCOUT_API_KEY

Or in ~/.codex/config.toml

[mcp_servers.scout]
url = "https://api.runscout.dev/mcp/your-team"
bearer_token_env_var = "SCOUT_API_KEY"

Add one line to your agent's instruction file, so it reads Scout before anything else and goes with Scout when the two disagree. Models give that file more weight than anything a tool returns.

At the start of every session, call scout_session_bundle and read it before you plan or change anything. When this file or the code disagrees with Scout's current decision, follow Scout's and say so.

Where it goes: ~/.claude/CLAUDE.md for Claude Code, ~/.codex/AGENTS.md for Codex, and an always-applied rule in .cursor/rules for Cursor.

VS Code

.vscode/mcp.json, or MCP: Open User Configuration

{
  "servers": {
    "scout": {
      "type": "http",
      "url": "https://api.runscout.dev/mcp/your-team",
      "headers": { "Authorization": "Bearer ${env:SCOUT_API_KEY}" }
    }
  }
}

Add one line to your agent's instruction file, so it reads Scout before anything else and goes with Scout when the two disagree. Models give that file more weight than anything a tool returns.

At the start of every session, call scout_session_bundle and read it before you plan or change anything. When this file or the code disagrees with Scout's current decision, follow Scout's and say so.

Where it goes: ~/.claude/CLAUDE.md for Claude Code, ~/.codex/AGENTS.md for Codex, and an always-applied rule in .cursor/rules for Cursor.

Grok Build

Terminal, or ~/.grok/config.toml

In a terminal

grok mcp add --transport http scout https://api.runscout.dev/mcp/your-team --header "Authorization: Bearer ${SCOUT_API_KEY}"

Or in ~/.grok/config.toml

[mcp_servers.scout]
url = "https://api.runscout.dev/mcp/your-team"
bearer_token_env_var = "SCOUT_API_KEY"

Grok Build connects with your API key, so there is no browser sign-in on this tab. Using Grok Bot? It connects with no key — see that tab. Run the command in a terminal where SCOUT_API_KEY is set, or use the config, which reads that variable when Grok starts. Leave off --scope project so this stays in your user config, not in the repository.

Add one line to your agent's instruction file, so it reads Scout before anything else and goes with Scout when the two disagree. Models give that file more weight than anything a tool returns.

At the start of every session, call scout_session_bundle and read it before you plan or change anything. When this file or the code disagrees with Scout's current decision, follow Scout's and say so.

Where it goes: ~/.claude/CLAUDE.md for Claude Code, ~/.codex/AGENTS.md for Codex, and an always-applied rule in .cursor/rules for Cursor.

Grok Bot

Your bot's chat in the Grok Bot app, not Slack

Paste this to your bot

Add the Scout MCP server at https://api.runscout.dev/mcp/your-team as a connector.

No API key. Your bot adds Scout from the MCP URL alone, then shows a Scout sign-in card in the chat. Approve it, and Scout's tools load on your bot's next message. The card only appears in your bot's chat in the Grok Bot app. Ask from Slack and it never shows.

Add one line to your agent's instruction file, so it reads Scout before anything else and goes with Scout when the two disagree. Models give that file more weight than anything a tool returns.

At the start of every session, call scout_session_bundle and read it before you plan or change anything. When this file or the code disagrees with Scout's current decision, follow Scout's and say so.

Where it goes: ~/.claude/CLAUDE.md for Claude Code, ~/.codex/AGENTS.md for Codex, and an always-applied rule in .cursor/rules for Cursor.

Other

Most MCP clients take this shape

{
  "mcpServers": {
    "scout": {
      "url": "https://api.runscout.dev/mcp/your-team",
      "headers": { "Authorization": "Bearer ${SCOUT_API_KEY}" }
    }
  }
}

Use your client's own syntax for reading an environment variable. The key-based steps read SCOUT_API_KEY. If your client cannot send a header, sign in through the browser against the MCP URL instead, where the client offers that.

Add one line to your agent's instruction file, so it reads Scout before anything else and goes with Scout when the two disagree. Models give that file more weight than anything a tool returns.

At the start of every session, call scout_session_bundle and read it before you plan or change anything. When this file or the code disagrees with Scout's current decision, follow Scout's and say so.

Where it goes: ~/.claude/CLAUDE.md for Claude Code, ~/.codex/AGENTS.md for Codex, and an always-applied rule in .cursor/rules for Cursor.

READ NEXT