№ 0219Skill
Deployable Muse connector template
An Apache-2.0 template for Meta Muse connectors with per-connection tokens, async job scaffolding and a single registry that generates routes, OpenAPI and llms.txt, deployable to Vercel, Workers or Docker.
Muse Connector Template
English · 简体中文
A deployable connector template for Meta Muse. Fork it, rename
it, point it at your own service. Apache-2.0, no strings.
📖 Read the step-by-step guide ·
中文文档
Muse is Meta's personal AI agent. It reaches third-party services through
connectors. Building one needs no app registration and no app id — you
expose an HTTP API that Muse can read about and call. This is that API, with the
parts that are easy to get wrong already done.
This deploys a working skeleton, not a finished connector. The operations
are placeholders. What you get is the auth model, the async-job plumbing, the
OpenAPI/llms.txt generation and the deploy wiring — so adding a real
capability becomes a small, local change.
Not locked in
Worth stating plainly, because connector platforms usually are:
- Deploy anywhere. Vercel, Cloudflare Workers, Docker, a VPS, Fly, Railway.
Same source, three adapters, no platform-specific code in the core. - Storage is a port, not a vendor. In-memory, Cloudflare KV, or any
Upstash-compatible REST endpoint. Implement the four-methodConnectionStore
interface and use Postgres, DynamoDB, or a file if you prefer. - No account with us, no API key from us, no service to sign up for.
CONNECTOR_SECRETis a string you generate. There is nothing to call home and
no telemetry. - The output is a plain OpenAPI service. Muse reads it, but so can anything
else that speaks OpenAPI — ChatGPT, C



ChatForm
Tgmlabs