Sharing one memory across Claude Code, Codex and local models over MCP: what worked
JoeBro121211 · reddit · 2026-11-23
A developer open-sourced a Mac app (GPLv3) that gives multiple coding agents and local models a shared per-project memory: one SQLite file exposed as an MCP server with remember/recall/supersede/forget tools, plus a project brief injected at session start.
Key design lessons:
- Provenance: every fact records which agent/model/machine wrote it, so recalled facts can be weighted by source
- Decay: never-recalled facts fade out, preventing recall quality from degrading
- Explicit supersede: replacing a fact is its own operation, avoiding stale versions winning retrieval
- Small facts beat summaries: one decision or gotcha per fact recalls far better
- Local models as first-class: Ollama/LM Studio/llama.cpp models share the same tools, so cheap local models can build on frontier-model work offline
Repo: joexk1/JoeBro. Open questions include decay aggressiveness and lexical vs. hybrid retrieval.
More from coding & agent
- huashu-design 3.5 ships: WeChat mini-program design support, two days after 3.4 — AlchainHust · 2026-11-23
- Dev Fixes 3 Agent Eval Holes: Blind Recount Hits 89% Agreement, Pooled 4% Miss Rate Hid a Bad Lane — armanidev_ · 2026-11-23
- Firecrawl vs context.dev: Developers weigh web data stacks for AI agents — lewishorsley96 · 2026-11-23
- Popfold: an open-source three.js engine built with parallel Claude Code sessions — cherichita · 2026-11-23
- Weaviate Podcast #150 dives deep into rerankers and agentic search — CShorten30 · 2026-11-23
- Job-hunting agent fixes reviewed: idempotency keys help, but AUC stays 0.645 — nivokvo · 2026-11-23