Memgraph IngesterStructure-aware RAG for code and project knowledge

Make agents smarter, index local source files into Memgraph so AI agents can discover symbols, follow relationships, search semantic code chunks, and keep durable project memory.

Get Started

GitHub

Make agents smarter immediately

Gives AI agents graph-backed code context, semantic retrieval, and durable project memory before the first investigation begins.

Graph-first code context

Creates project-scoped File, Package, Class, Interface, Annotation, Method, and Field nodes with source locations and relationships.

RAG that verifies

Uses derived CodeChunk and MemoryChunk search nodes, then links retrieval back to canonical graph records and source files.

Local by default

Reads files locally and writes to your Memgraph instance over Bolt. Source code is not uploaded by the ingester.

Multi-language ingestion

Supports Java, JavaScript, TypeScript, Python, and ctags-detected fallback languages for structural inventories.

Agent instructions

Can install Codex, Claude, Gemini, GitHub Copilot, or raw Cypher guidance into your project instruction files.

Durable memory

Optional project memory stores rules, decisions, findings, tasks, risks, questions, and code references beside the code graph.

The Short Version

Memgraph Ingester turns a repository into a queryable code graph for AI-assisted development. Instead of making an agent repeatedly scan files, you can give it indexed structure, semantic retrieval, source ranges, call edges, type relationships, and project memories.

bash

docker run -p 7687:7687 -p 7444:7444 --name memgraph memgraph/memgraph-mage:3.9.0

memgraph-ingester \
  --source . \
  --bolt bolt://localhost:7687 \
  --project my-project \
  --wipe-project-code \
  --init-instructions \
  --apply-schema

Then connect your agent through memgraph-ingester-mcp and ask questions about the codebase through graph-aware tools.