oomol-lab/open-connector — explained in plain English
Analysis updated 2026-05-18
Give an AI agent safe access to a user's GitHub, Gmail, or Slack account without exposing their credentials.
Build an agent product that needs reusable, permissioned access to many different SaaS tools.
Self host a connector gateway locally, on Fly.io, or on Cloudflare Workers instead of using a hosted service.
Expose a catalog of prebuilt Actions to agent hosts through the Model Context Protocol.
| oomol-lab/open-connector | colbymchenry/codegraph | dakheera47/job-ops | |
|---|---|---|---|
| Stars | 3,095 | 3,101 | 3,124 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | easy | moderate |
| Complexity | 4/5 | 2/5 | 3/5 |
| Audience | developer | developer | general |
Figures from each repo's GitHub metadata at analysis time.
Requires Node.js 22 or newer and choosing a deployment target such as Docker, Fly.io, or Cloudflare Workers.
OpenConnector is an open source gateway that lets AI agents and apps connect to over a thousand different online services, such as GitHub, Gmail, Notion, BigQuery, Google Analytics, Supabase, Airtable, and Slack, without giving the agent direct access to a user's login credentials. Instead of building a separate connection for every service an agent needs, a developer connects a user's account once through OpenConnector, and the agent can then call any of over ten thousand prebuilt Actions against that account through a shared catalog. The gateway sits between the agent and the outside service, holding onto API keys, OAuth logins, and other credentials in a runtime that keeps a record of what was done. Every Action has a defined request and response shape, a list of required permissions, and can be turned on or off through allow and block policies. Sensitive details in run logs are hidden, and temporary files used during a run are handled separately from permanent credential storage. Developers can reach OpenConnector in several ways depending on what they are building: a TypeScript SDK for calling it from app code, a command line tool called oo CLI that acts as a local relay for an agent, the Model Context Protocol for agent hosts that support it, or plain HTTP and OpenAPI calls for custom clients. A web console is included for browsing connected services, setting up credentials, creating runtime tokens, and watching recent activity and failures. The project can run in several ways: self hosted locally with Docker or Node.js and a SQLite database, hosted on Fly.io with persistent storage, deployed to Cloudflare Workers using D1 and R2 for storage, or used through OOMOL's own hosted version of the same system for teams that are not ready to run their own infrastructure. Whichever option is used, the provider ids, Action ids, and data shapes stay the same, so a project can start on the hosted version and move to self hosting later without rewriting how it talks to the gateway. The project is licensed under Apache 2.0.
An open source gateway that lets AI agents securely connect to over 1,000 online services like GitHub, Gmail, and Slack without handing over user credentials.
Mainly TypeScript. The stack also includes TypeScript, Node.js, Cloudflare Workers.
Use freely for any purpose, including commercial use, as long as you keep the required copyright and license notices.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.