langchain-ai/chat-langchain

A documentation assistant deployed as a Managed Deep Agent.

This is a documentation assistant agent that helps answer questions about LangChain, LangGraph, and LangSmith. It demonstrates how to build a production-ready agent using:

• Managed Deep Agents - For managed deployment, identity, and connectors
• LangChain Agents - For agent creation with middleware support
• Guardrails - To keep conversations on-topic

The repo also includes a Next.js frontend in frontend/ for the public chat UI.

• Documentation Search - Searches official LangChain docs via managed MCP
• Support KB - Searches the Pylon knowledge base for known issues
• Link Validation - Verifies URLs before including in responses
• Guardrails - Filters off-topic queries

• Python 3.11+
• uv (recommended) or pip

# Clone the repository
git clone https://github.com/langchain-ai/chat-langchain.git
cd chat-langchain

# Install dependencies with uv
uv sync

# Or with pip
pip install -e .

# Copy environment template
cp .env.example .env

# Edit .env with your API keys

# Build the Managed Deep Agent bundle
uv run mda dev .

# Or with pip
mda dev .

cd frontend
npm ci
npm run dev:local

Point the frontend at the local MDA deployment via NEXT_PUBLIC_LANGGRAPH_API_URL (see frontend/.env.local.example). Signed-in and anonymous users authenticate with Supabase; MDA validates those tokens and scopes their threads.

├── agent.py                    # Managed Deep Agent entrypoint
├── identity.py                 # MDA multi-region Supabase identity contract
├── instructions.md             # Managed Deep Agent system prompt
├── tools/
│   └── mcp.py                  # Managed MCP docs declaration
├── src/
│   ├── agent/
│   │   └── config.py           # Model configuration
│   ├── tools/
│   │   ├── pylon_tools.py      # Support KB tools
│   │   ├── pricing_tools.py    # Pricing fetch
│   │   └── link_check_tools.py # URL validation
│   ├── prompts/
│   │   ├── docs_agent_prompt.py # Hub push / eval mirror of instructions.md
│   │   ├── guardrails_prompts.py
│   │   └── context_summary_prompt.py
│   └── middleware/
│       ├── guardrails_middleware.py
│       ├── ingress_guards_middleware.py
│       └── retry_middleware.py
├── frontend/                   # Next.js public chat UI
└── pyproject.toml              # Python project config

The agent uses a docs-first research strategy:

1. Guardrails Check - Validates the query is LangChain-related
2. Documentation Search - Searches official docs via the managed MCP connector
3. Knowledge Base - Searches Pylon for known issues/solutions
4. Link Validation - Verifies any URLs before including them
5. Response Generation - Synthesizes a helpful answer

mda deploy .

What MDA owns in this deployment:

• Identity — identity.py verifies signed-in and anonymous Supabase access tokens across the configured regions.
• HTTP surface — managed agent ingress plus same-origin frontend routes.
• LangSmith browser ops — the frontend server verifies MDA thread access, then proxies feedback and trace read/share without exposing LANGSMITH_API_KEY.
• Docs MCP — tools/mcp.py attaches the LangChain docs MCP tools.
• Thread titles — generated in the browser (deterministic truncation).
• Checkpointer — managed by the Managed Deep Agents runtime.

• LangChain Documentation
• LangGraph Documentation
• LangSmith Documentation

MIT