# Krusch-Git MCP server

> Indexes Git repositories into a semantic graph using PostgreSQL and pgvector to enable AST-based symbol search and hybrid code retrieval.

- Repository: https://github.com/kruschdev/krusch-git
- Page: https://hostedmcp.org/servers/kruschdev/krusch-git
- Transport: —
- Stars: 0
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `DATABASE_URL` | no | no |  |
| `EMBED_MODEL` | no | no |  |
| `OLLAMA_URL` | no | no |  |

## Tools (7)

- `krusch_git_list_repos`
- `krusch_git_read_tree`
- `krusch_git_read_blob`
- `krusch_git_semantic_search`
- `krusch_git_search_symbols`
- `krusch_git_file_symbols`
- `krusch_git_dependency_graph`

## FAQ

### Does Krusch-Git replace Git?

No, it is an agentic augmentation layer that sits alongside Git to provide a semantically searchable clone of your codebase for AI agents.

### What databases are required for Krusch-Git?

You need a running PostgreSQL instance with the pgvector extension enabled.

### How does the server handle large files and memory usage?

It uses a pointer-mode storage system where file content is stored as pointers to the local filesystem, reducing database overhead by over 85%.

### What embedding models are supported?

It works with Ollama-hosted models, defaulting to snowflake-arctic-embed2, but also supports qwen3-embedding and bge-large.