About
About Cymatix Context
Cymatix Context is an open-source, local-first context engine for LLM
agents, released under the Apache-2.0 license and published on PyPI as
cymatix-context. This site, cymatixcontext.com, is the
project's official home page and wiki.
What it does
- Keeps a persistent knowledge store of your codebase and documents in a single SQLite file on your own machine
- Runs a seven-stage retrieval pipeline that assembles a token-budgeted context window for each agent turn, with no model inference on the default retrieval path
- Answers every request with an explicit know/miss contract, so an agent can tell grounded context from a miss and knows when to reread or escalate
- Reaches agents and editors four ways over the same JSON shapes: a CLI, an MCP server, an OpenAI-compatible proxy, and a plain HTTP API
Who builds it
- Created and maintained by Michael Bachaud (@mbachaud on GitHub)
- Developed in the open at github.com/mbachaud/Cymatix-Context; outside contributions are welcome under the contributor guide
- The background writing behind the design is Agentome
How the project works
- Defaults change only behind a published measurement, and costly results are disclosed too: see Benchmarks and Receipts
- Releases, highlights, and the deferred ledger are on Roadmap and Releases
- The software runs wherever Python runs; Hosting lists self-hosting and a managed option
- Questions and reports: Contact