🧠 AgensGraph AI Tools and Libraries

Welcome to AgensGraph AI — a curated collection of tools, integrations, and starter templates for building AI-powered applications that work with AgensGraph, a PostgreSQL-based multi-model graph database.

This repository currently includes:

• LangChain — LLM agents, tools, and chains
• LlamaIndex — custom data indexing and retrieval
• LightRag — graph-aware RAG for accurate, context-rich retrieval
• cognee — dynamic memory for Agents
• mcp — Model Context Protocol server for AgensGraph enabling database access and graph exploration.

✅ Each library has its own subfolder with a dedicated README to guide you through setup and usage.

📦 Installation

The agensgraph-ai package installs any combination of the integrations under one name. Pick the ones you need:

pip install "agensgraph-ai[langchain]"
pip install "agensgraph-ai[langchain,lightrag]"
pip install "agensgraph-ai[all]"

Name at least one extra. pip install agensgraph-ai on its own installs no integrations, and an extra that is misspelled installs none either — pip warns about that, uv does not.

Each integration is also released on its own, so it can be installed by name instead. The two forms produce the same environment — agensgraph-ai ships no code, and is a convenience rather than a layer:

pip install langchain-agensgraph
pip install llama-index-agensgraph
pip install lightrag-agensgraph
pip install cognee-agensgraph

The MCP servers are commands rather than libraries, and an MCP client normally launches them itself with uvx, which needs no install at all:

"mcpServers": {
  "agensgraph-cypher": {
    "command": "uvx",
    "args": ["mcp-agensgraph-cypher@0.3.0", "--transport", "stdio"]
  }
}

The mcp extra is for the other case: hosting a server yourself over HTTP or SSE.

Database requirements

The Python install is only half of the setup. These integrations talk to a running AgensGraph, 2.17 or newer, and the vector-backed features need the pgvector and meta extensions, which AgensGraph does not bundle — see langchain/README.md for how to build and enable them. SHOW agversion tells you which you have. From 2.18 a release reports four numbers, such as 2.18.4.0, with -rc1 on a release candidate; the first two are the line, so 2.18.6.0-rc1 is a 2.18 server.

🎯 Purpose

This repository is designed to help developers:

• Integrate AgensGraph with modern LLM frameworks
• Leverage graph data in conversational and intelligent apps
• Explore Retrieval-Augmented Generation (RAG), agents, and graph reasoning

Everything is open-source and modular — feel free to use, fork, or contribute.

📄 License

This repository is licensed under the Apache License 2.0.

📬 Contact

For questions, feature requests, or collaboration:

• Open an Issue or Pull Request