DoubleTrust - Agents discovery & governance platform

A platform for discovering and governing your Agents inside your codebases.

Quick Start

1. Setup:

   # Install Python dependencies
   pip install -r requirements.txt
   
   # Install Node.js dependencies
   cd frontend && npm install && cd ..
   
   # Copy .env.example in a .env file and set your OpenRouter API key
   cp .env.example .env

2. Launch:

   # Start the platform
   python doubletrust.py

3. Access:

   • Frontend: http://localhost:3000
   • Backend API: http://localhost:8000/docs

Features

Agent Discovery

• GitHub Repository Scanning: Discovers agents from GitHub repos
• Framework Detection: Identifies LangChain agents (create_react_agent) and Custom agents
• Tool Extraction:
  • LangChain: Extracts tools from create_react_agent parameters
  • Custom: Uses LLM to detect tools from system prompts
• Risk Assessment: LLM-powered risk evaluation (low/medium/high) based on agent role and tools

Agent Management

• Framework Tags: Visual indicators for LangChain vs Custom agents
• Risk Indicators: Color-coded risk levels with hover explanations
• Tool Management: Per-agent tool lists with descriptions
• Database: SQLite with agents, agent_tools tables

Project Structure

doubletrust/
├── doubletrust.py                  # Unified launcher (backend + frontend)
├── backend/
│   ├── main.py                     # FastAPI application
│   ├── database.py                 # SQLite with agents/agent_tools tables
│   ├── models/agents.py            # Agent models with framework/risk fields
│   ├── services/
│   │   ├── discovery_service.py    # Orchestrates discovery + risk assessment
│   │   ├── github_service.py       # GitHub operations
│   │   └── discovery/
│   │       ├── discovery.py         # Main discovery logic
│   │       ├── extractor.py         # LangChain + Custom agent extraction
│   │       └── role_assigner.py    # Role assignment
│   └── api/
│       ├── agents.py               # Agent endpoints
│       └── discovery.py            # Discovery endpoints
├── frontend/
│   └── src/components/
│       ├── agents/                 # Agent cards with framework/risk tags
│       ├── discovery/              # GitHub repo discovery
│       └── layout/                 # Navigation (Agents, Discovery)
├── llm_service/
│   ├── llm.py                      # OpenRouter integration
│   └── prompts/
│       ├── tool_detection.py       # Tool extraction from prompts
│       └── agent_risk.py          # Risk assessment
└── agent-test/                     # Sample LangChain/Custom agents

Testing

Use the sample agents in agent-test/:

• lc_agent_inline.py - LangChain agent with inline tools
• lc_agent_assigned.py - LangChain agent with variable-assigned tools
• custom_agent.py - Custom agent with system prompt

Development

• Database: Auto-created SQLite doubletrust.db with migration support
• LLM Integration: OpenRouter API for tool detection and risk assessment
• Framework Detection: AST parsing for LangChain create_react_agent calls
• Risk Assessment: LLM evaluation based on agent role and discovered tools

More to come

• Support more frameworks (CrewAI, Langgraph...)
• MCP discovery