API & Architecture Overview

The Agent Bridge backend is built with FastAPI and follows a modular service-oriented architecture.

Core Modules

Module	Description
Agent Services	Handles the LLM orchestration, LangGraph workflows, and dynamic tool binding.
MCP Services	Manages connections to Model Context Protocol servers, including OAuth authentication and tool discovery.
Database Models	SQLAlchemy models for persistence (Users, Settings, Permissions).
Auth	JWT-based authentication and security utilities.

Data Flow

1. User Request: Client sends a chat message to POST /chat.
2. Agent Initialization: agent_factory builds a user-specific graph with tools relevant to the user's connected MCP servers.
3. Graph Execution: The LangGraph workflow (agent_orchestrator) executes.
   • Agent Node: LLM generates a response or tool call.
   • Permissions Check: If a tool is called, permissions are checked against tool_permissions.
   • Human Review: If required, execution pauses for user approval.
   • Tool Execution: Approved tools are executed via MCPConnector.
4. Response: Final answer is streamed back to the client.

Directory Structure

backend/app/
├── services/
│   ├── agent/       # LLM & Graph Logic
│   └── mcp/         # MCP Server Connections
├── models/          # Database Schema
├── routes/          # API Endpoints
└── auth/            # Security & JWT