Anil-matcha/open-agent-connector

Open-source, self-hosted alternative to Composio, Nango, Merge.dev, and Arcade.dev for AI agent integrations. Run GitHub, Slack, Hacker News, and custom HTTP actions over MCP or REST.

What it solves

Open Agent Connector provides a self-hosted gateway that allows AI agents to securely connect to and interact with third-party application APIs. It solves the problem of managing credentials and discovering available actions in a consistent way, allowing developers to keep their API keys and tokens within their own runtime rather than relying on managed third-party integration platforms.

How it works

The system consists of a FastAPI backend that manages provider definitions, encrypted credentials, and outbound requests, and a Next.js web console for managing connections. It exposes these capabilities to agents via two primary interfaces: the Model Context Protocol (MCP) and a standard REST API. When an agent wants to perform a task, it can discover available providers (like GitHub or Slack), inspect the required input schemas for specific actions, and execute those actions through the gateway.

Who it’s for

It is designed for developers and teams who want a self-hosted, extensible integration layer for their AI agents and prefer to own their integration service rather than using managed alternatives like Composio or Nango.

Highlights

  • MCP Gateway: Full support for the Model Context Protocol, allowing agents to discover and execute tools via JSON-RPC.
  • Secure Credential Storage: Connection credentials are encrypted at rest using AES-256-GCM.
  • Request Safeguards: Built-in protection against Server-Side Request Forgery (SSRF) by blocking private and internal network targets by default.
  • Idempotent Execution: Supports Idempotency-Key to prevent duplicate upstream API calls.
  • Extensible Catalog: Providers are defined in Python, making it easy for developers to add new services and actions.

Related

  • Project
    BradGroux/veritas-kanban

    A local-first Kanban board that scales from simple task management to a governed, multi-agent orchestration platform for autonomous coding workflows.

  • Project
    vercel-labs/ralph-loop-agent

    A framework for continuous AI agent autonomy that wraps the AI SDK in an outer loop to iteratively verify and complete complex tasks.

  • Project
    formulahendry/wechat-acp

    A bridge that connects WeChat direct messages to ACP-compatible AI agents, allowing users to interact with local AI CLI tools via WeChat.

  • Project
    RUC-NLPIR/Awesome-Long-Horizon-Agents

    A GitHub repository that curates 750+ papers, benchmarks, and resources about long-horizon AI agents, organized by themes like evolution, harness engineering, optimization, and applications. It accompanies a survey paper and serves as a map of the field.