Show HN: A local memory that Claude Code, Codex (and your fav agent) share
onadit.comWe're so fucking spoiled by the AI labs right now. Gemini 4 Carbon just dropped, 2 weeks after Argon. Opus 5.5 is still so good, Codex is ofc cooking something, maybe Grok will catch up!
When I open a fresh Codex session, I think like "wtf I explained all of this in claude code like an hour ago". When a lab ships memory inside its own app, I think: that's lock-in, not memory, why can't it be open to other agents? When I add another rule to CLAUDE.md, I think: half of what matters never makes it in there :D
that's why i deeply believe my memory should sit on my disk, and every agent should read from it.
well, that's Onadit :)
It files my Claude Code, Codex and Grok CLI sessions every hour. Any agent asks it over MCP and gets 6 KB of decisions and rules back, each with the session it came from, in 65 ms on a repeat question (yes, fast). No account, no telemetry, and your conversations never reach me. Search by meaning is opt-in (Voyage embeddings). Every network call is listed at onadit.com/docs/egress.
I use it all day. On a 16-task test over my own archive it finds the right source 7 times in 10. Ok that's my archive, not a benchmark, but still!
7 days free, then €5 a month. Closed source. Mac, Apple silicon.
onadit.com
No comments yet.