My agent could read any file I handed it and had no idea what else existed in the repo.
Ask it to change a function and it cannot tell you what calls that function, who wrote it, or that the same change was tried and reverted in March. So I became the index. I grepped, I pasted, I explained, and then I did it again the next session, because nothing I told it yesterday survived.
I built Oynix to stop doing that. One command indexes your codebase on your machine, and the agent queries it directly over MCP. What does this symbol do. What calls it. Who owns it. What breaks if I change it. What did the last session already try. It can also write down what it just did, so the next session, or a different agent
entirely, reads that back instead of starting from nothing.
Your AI coding agent knows your code but not your company. Oynix gives it organizational memory, not just code search. It builds one live graph across repos, docs, Jira, ownership, incidents and past decisions, and serves it over MCP inside agnets like Claude Code, Codex and 16 others. Now agents answer cross-repo "why does this exist / what breaks / who owns this" from real company memory, on your own keys, instead of starting from zero every session.
