A containerized implementation of Graphify exposing a Model Context Protocol (MCP) server over HTTP/SSE. This image is designed to be easily consumed by autonomous agents (like Hermes) via Docker Compose.
- MCP Ready: Exposes Graphify's API over HTTP (SSE) for seamless agent integration.
- Multi-LLM Support: Pre-configured environment variables for Ollama, Anthropic, Gemini, and OpenAI.
- IPv6 Enabled: Configured for modern networking architectures with IPv6 support.
- Immutable & Deterministic: Built directly with
uv tool, meaninggraphifyy[all]is pre-baked into the image. No downloads at runtime.
The MCP server reads from a static graph file. Before starting the server, you must generate the graph.json file from your project code.
Run this locally (or inside an ephemeral container) in your project directory (e.g., ./app):
uvx graphifyy extractThis will create a graphify-out/ directory containing the required data. Make sure this folder is located inside the ./app directory mapped in your docker-compose.yml.
Use the provided docker-compose.yml to bring up the service. The container will automatically use the pre-installed binary to serve your graph.
docker compose up -dConfigure your agent (e.g., Hermes) to connect to the remote MCP server using the SSE transport. If both containers share the same Docker network, use the internal hostname:
- Endpoint URL:
http://graphify:8080/sse
Modify the environment section in your docker-compose.yml to configure your preferred LLM provider.
If you are running Ollama on your Docker host machine, be sure to use host.docker.internal or your host's local IP address instead of localhost:
OLLAMA_BASE_URL=[http://host.docker.internal:11434](http://host.docker.internal:11434)
GRAPHIFY_OLLAMA_KEEP_ALIVE=10Uncomment and populate the relevant variables for cloud providers:
ANTHROPIC_API_KEY/ANTHROPIC_MODELGEMINI_API_KEYOPENAI_API_KEY/OPENAI_BASE_URL/OPENAI_MODEL
Docker Hub: sinfallas/remote-graphify