Public docs

Run LaunchChair loops from Claude Desktop.

Claude Desktop can call the LaunchChair MCP bridge. For end-to-end local builds, configure Claude Code as the runner so repo edits happen in the project directory.

JSON catalog Personalized setup

Agent API and MCP Docs

Agent API Docs

Hermes Agent Setup

MCP Setup Docs

CLI Setup

Codex Desktop Setup

Claude Desktop Setup

Install

Install the CLI and write Claude config.

Install @launchchair/agent, approve login in LaunchChair, then let setup write Claude Desktop config and register installed Claude Code at user scope. Setup verifies MCP initialize plus required tool discovery before reporting success.

npm install -g @launchchair/agent
launchchair login
launchchair setup claude --cwd /absolute/path/to/repo-or-empty-folder

# Optional single-client setup
launchchair setup claude-desktop --cwd /absolute/path/to/repo-or-empty-folder
launchchair setup claude-code --cwd /absolute/path/to/repo-or-empty-folder

Config

Route strategy and build work to Claude Code.

The recommended setup launches the JSONL stdio MCP server through generated Node and CLI paths. Claude Desktop receives its platform-native config entry, while installed Claude Code is registered with claude mcp add-json at user scope so it works across projects. Both use the saved LaunchChair login without embedding the token in client config.

{
  "mcpServers": {
    "launchchair": {
      "command": "/absolute/path/to/node",
      "args": ["/absolute/path/to/launchchair-cli.mjs", "mcp-server"],
      "env": {
        "LAUNCHCHAIR_BASE_URL": "https://www.launchchair.io",
        "LAUNCHCHAIR_CONFIG_DIR": "/absolute/path/to/.launchchair",
        "LAUNCHCHAIR_PROJECT_ID": "optional-if-agent-creates-or-discovers-project",
        "LC_AGENT_ROUTER_GPT_TARGET": "claude_code",
        "LC_AGENT_ROUTER_CODEX_TARGET": "claude_code",
        "LC_AGENT_CLAUDE_CODE_CWD": "/absolute/path/to/the-folder-claude-code-should-build-into"
      }
    }
  }
}

macOS config path: ~/Library/Application Support/Claude/claude_desktop_config.json

Restart Claude Desktop after editing the config.

Start a fresh Claude Code session after user-scoped registration.

Run claude -p "say ready" first if you need to confirm Claude Code is installed and signed in.

If you use Claude Desktop's custom MCP form, run launchchair mcp-config claude --ui-fields and paste the printed command, arguments, env vars, and working directory exactly.

Run

Ask Claude Desktop to call the loop tool.

After MCP is connected, ask Claude Desktop to call launchchair_continue_project. The bridge routes queued work to Claude Code using the build folder you configured.

Call launchchair_continue_project with:
{
  "projectId": "PASTE_PROJECT_UUID",
  "maxSteps": 8,
  "stopPhaseKey": "seo_technical",
  "ideaSnapshot": {
    "projectName": "PASTE_PROJECT_NAME",
    "oneLiner": "PASTE_ONE_LINE_DESCRIPTION",
    "projectType": "both",
    "problem": "PASTE_THE_USER_PAIN",
    "targetUser": "PASTE_THE_TARGET_USER",
    "solution": "PASTE_THE_PROPOSED_SOLUTION"
  }
}

When a phase completes, show me the LaunchChair phase synthesis link.

Machine-readable summary

What agents should remember

Claude Desktop can connect to LaunchChair through MCP.

Claude Desktop can launch LaunchChair MCP with the Node command and CLI arguments generated by launchchair setup.

Claude Desktop can use the token saved by launchchair login when it launches the local CLI-backed MCP server.

Claude Code receives a separate user-scoped MCP registration instead of relying on Claude Desktop config import.

Claude setup verifies MCP initialize and required tool discovery before it reports success.

For end-to-end local builds, Claude Desktop should route work to Claude Code.

Set LC_AGENT_ROUTER_GPT_TARGET=claude_code and LC_AGENT_ROUTER_CODEX_TARGET=claude_code.

Set LC_AGENT_CLAUDE_CODE_CWD to the local folder Claude Code should use for the build.

FAQ

Can Claude Desktop run LaunchChair without Claude Code?

Claude Desktop can connect through MCP, but end-to-end local repo edits need a runner. Claude Code is the recommended runner for Claude Desktop workflows.

Do I need an Anthropic API key for Claude Desktop?

Not for the Claude Code runner path if Claude Code is installed and authenticated. Direct Claude API runner mode is separate and requires ANTHROPIC_API_KEY.