This service provides an isolated PostgreSQL database with vector capabilities (pgvector) and a graph-based code indexing engine. It allows you to transform codebases into searchable graphs and expose them as context to AI coding agents via the Model Context Protocol (MCP).

What it does

…

1. Indexes Codebases: Maps your files, entities, and relationships into a searchable graph.
2. Stores Vectors: Prepares your codebase for vector-based semantic search.
3. Serves MCP: Provides the graph to AI agents (like Claude CLI or Gemini CLI) so they can understand the structure of your projects.

Architecture at a glance

   host codebase                  docker compose stack
   -------------                  --------------------

  /path/to/project  --(ro)-->  +-----------+
                               |  graphify |  one-shot indexing job
                               +-----------+
                                     |  writes nodes and edges
                                     v
                               +-----------+
                               | postgres  |  pgvector/pgvector:pg16
                               +-----------+
                                ^    ^    ^  read
                                |    |    |
                     +-----------+  +--------+  +-----+
   Claude, Gemini <->| mcp-server|  | viewer |  | web |<-> browser
                     +-----------+  +--------+  +-----+

• postgres: Stores the graph, vector embeddings, and persistent plans.
• graphify: Analyzes your codebase and populates the database.
• mcp-server: Exposes the graph data via the Model Context Protocol.
• viewer: Renders an interactive visualization of your codebase.
• web: A management dashboard for indexed projects and plans.