agent_bridge

Purpose

Agent Bridge is a modular AI agent platform for connecting LangChain, FastAPI, and Model Context Protocol (MCP) servers. Its goal is to provide an end-to-end, deployable, production-grade framework for remote MCP servers, helping developers break free from the limitations of local stdio isolation and build and deploy enterprise-grade agents. The project has been officially included in the Awesome MCP Clients repository.

Main Capabilities

• AI Agent Orchestration: Integrates LangChain and LangGraph state machines on top of a FastAPI backend.
• Multi-Server MCP Support: Can register, manage, and execute tools across multiple Model Context Protocol servers.
• Authentication and Security: JWT user management, bcrypt password hashing, and OAuth 2.0 with PKCE support (examples include Figma, Notion, GitHub).
• Human-in-the-Loop Permissions: Provides fine-grained tool execution control and permission requests.
• Modern Web Interface: A responsive frontend based on Vue 3, Vite, Tailwind CSS, and Element Plus components.
• Real-Time Streaming: Streams agent interactions through SSE (Server-Sent Events) and WebSocket pipelines.
• Flexible Database Layer: PostgreSQL database using SQLAlchemy ORM, with SQLite support for local development.

Usage

Docker Quick Start

1. Clone the repository:git clone https://github.com/ramblinghermit0403/agent_bridge.git, then cd agent_bridge.
2. Configure environment variables: create a backend/(or server/, depending on the specific setup) .env file:cp backend/.env.example backend/.env, and fill in the API credentials (such as GOOGLE_API_KEY、OPENAI_API_KEY).
3. Start the services:docker-compose up --build.
   • Frontend UI:http://localhost:80
   • Backend API:http://localhost:8001
   • PostgreSQL: localhost:5432

Manual Installation (Development)

• Backend setup: the backend API will run at http://localhost:8001。
• Frontend setup: the web interface will run at http://localhost:5173。
• Documentation setup: the local documentation server runs at http://localhost:5173（或下一个可用端口）。

Configuration

• MCP Server Registration: To connect a custom MCP server, update backend/servers.json and run the bootstrap script.
• Database Migration: Database tables are automatically generated when the backend starts; SQLAlchemy migrations can be used for custom schema changes.

Use Cases

• Full Platform Deployment: For engineering teams, startups, and enterprises.
• Learning and Education: For developers learning Agentic AI and MCP.
• Modular Extraction: For product developers.
• MCP Server Testing and Sandboxing: For tool developers and authors.

Tech Stack and Security Governance

The project is built on FastAPI and LangChain and follows Anthropic's Model Context Protocol standard. Security aspects include: encrypted JWT session tokens and secure password hashing, CORS isolation and environment configuration control, role-based tool authorization, and real-time execution approval requests.

Limitations

Official materials do not provide specific performance metrics, charging models, prices, platform compatibility details, or effectiveness data; this information is not currently available. The project uses the MIT open source license, and contributions are welcome through CONTRIBUTING.md.