# mache MCP server

> Projects structured data like code, JSON, and YAML into a navigable graph for AI agents to explore via MCP tools or a mounted filesystem.

- Repository: https://github.com/agentic-research/mache
- Page: https://hostedmcp.org/servers/agentic-research/mache
- Transport: —
- Stars: 46
- License: Apache-2.0

## FAQ

### What kind of data can mache project?

Mache can project various structured sources including JSON documents, SQLite databases, and source code like Go or Rust, as well as Terraform, Markdown, and audit trails.

### Does mache require any other software to work?

Yes, ley-line-open is a hard requirement because mache uses it to produce the _ast database that all source projections read.

### How does an agent interact with the projected data?

Agents can interact with the data through 19 MCP tools that allow for navigation, tracing reference flows, judging structural quality, and editing files.

### Can I use mache to edit my code?

Yes, mache provides a `write_file` tool that allows for editing through a process of validation, formatting, and proposing changes.
