Axon
Indexes any codebase into a structural knowledge graph (dependency resolution, call-chain tracing, dead-code detection) and exposes it over MCP so AI coding agents can query the graph instead of re-reading files.
Axon started from token anxiety. Every coding agent I used burned tokens re-doing the same searches: grepping the same files, re-discovering the same structure, request after request. Context is scarce and expensive, and I kept hitting limits; it bothered me that the model kept paying to learn what it already knew a minute ago.
I gave it a graph instead. Axon parses source into nodes (files, modules, symbols) and edges (imports, calls, references), resolves dependencies across languages, and serves the whole thing as an MCP server plus a web dashboard. Any agent that speaks MCP (Claude Code, Cursor, Copilot) can ask "who calls this, what's dead, which module is load-bearing" and get the answer directly, without re-reading the repo.
I sat on it for a while, unsure anyone else would want it. When I finally open-sourced it, people did: DMs about how it was saving them tokens, feature requests, VCs reaching out and connecting me with companies that wanted to integrate it. That reaction is the part I'm proudest of; it went from a personal itch to something other people build on.
700+ stars and climbing, shipped on GitHub and installable with a single `pip install axoniq`. Still adding languages.