MEX uses Git itself as a shared context layer for coding agents and teams
DJIRNMAN · reddit · 2026-09-09
The author open-sourced MEX (mex-memory/mex), a Git-native shared memory layer for coding agents and engineering teams.
Core idea: what team memory needs — history, diffs, changes moving with the repo, per-branch state, visible conflicts — Git already does, so skip the database/sync/accounts stack.
- Memory as files: architecture, decisions, specs, workstreams, handoffs, and activity get committed like code
- Local indexes stay local: each checkout rebuilds its own search and code-intelligence indexes
- Memory shares code history: decision changes are reviewable via diff, conflicting edits surface as Git conflicts, and agent learnings survive the session
- Structured handoffs: agents prepare handoffs with completed work, blockers, decisions, changed files, and next actions, committed into project history
- No silent truth: shared changes go through a proposal flow — publishing and accepting a Spec are separate, explicitly approved actions
Tradeoffs: Git isn't a realtime bus, MEX doesn't auto push/pull, and disconnected clones still produce normal conflicts. The author argues project memory should behave more like source code than another SaaS database.
Related event: MEX: Using Git as Shared Memory Layer for Coding Agents(2 posts)→
More from coding & agent
- Agent Field Kit launches $29 security runbooks for production AI agents — alvelda · 2026-10-27
- Dev mixes Claude specs with abliterated GLM to build SS7 threat detector — l4rz · 2026-10-27
- Cloudflare Durable Objects validated as top architecture for AI agent WebSocket servers — tristanbob · 2026-10-27
- GitHub Next's Chopin prototype explores real-time multiplayer agentic planning — max__drake · 2026-10-27
- Veynod: An Open-Source Shared Workspace Where AI Agents Collaborate Like a Team — TKOWNS · 2026-10-27
- TMLR paper: enforced thinking doesn't reliably make LLM agents disclose key info to users — SharonYixuanLi · 2026-10-27