a day ago
LaunchChair Agent Loop MCP connects desktop agents like Codex and Claude Code to LaunchChair’s Agent API. Agents can create and find LaunchChair projects, run structured product ideation, validation, MVP blueprint, stack setup, and build-card loops, claim “Run with agent” jobs, apply structured outputs, retry or remediate failed work, and return links back to LaunchChair synthesis pages while LaunchChair remains the source of truth for the living spec.
Server Config
{
"mcpServers": {
"launchchair": {
"command": "node",
"args": [
"/Users/you/launchchair-agent/mcp-bridge-server.mjs"
],
"env": {
"LAUNCHCHAIR_BASE_URL": "https://www.launchchair.io",
"LAUNCHCHAIR_AGENT_TOKEN": "<YOUR_LAUNCHCHAIR_AGENT_TOKEN>",
"LAUNCHCHAIR_PROJECT_ID": "<OPTIONAL_PROJECT_UUID>",
"LC_AGENT_ROUTER_GPT_TARGET": "codex",
"LC_AGENT_ROUTER_CODEX_TARGET": "codex",
"LC_AGENT_CODEX_CWD": "/absolute/path/to/your/project-or-empty-build-folder"
}
}
}
}