GitHub for AI Agent Memory - Shared, versioned memory for your agents and your team

Every AI agent starts a new session with no memory of what it did before. SenseLab fixes that. Connect your agents via MCP or SDK, and they get persistent, versioned memory across sessions. Invite teammates into a shared Room where agents from different users read and write the same memory, with full provenance on who knew what and when. Works with Claude Code, Cursor, CrewAI, LangGraph, and AutoGen. Two env vars. Free to start.

Hey Product Hunt 👋 I'm Bruno, co-founder of SenseLab. I built SenseLab because agent memory for teams is completely broken, and everyone is just pretending it's fine. Your agents finish a session and forget everything. Ok, annoying but manageable. Then you add more agents, or more teammates running their own agents, and it falls apart completely. Agent A fixes a bug. Agent B doesn't know. Agent B fixes the same bug. Nobody knows what anyone else figured out. Every agent, every session, starts from zero. So teams reach for markdown files. Which sounds reasonable until your CLAUDE.md is 400 lines long, costs you $800/month in tokens just from file overhead, and still has no versioning, no shared state across agents, and no way to know which agent wrote what or why. Vector DBs? Great for retrieval, not built for agents writing and updating structured knowledge over time. So we built SenseLab. Persistent, versioned memory that works across every agent, every session, every teammate. Every read and write is tracked, so you always know which agent knew what and when. Rooms let you invite teammates so their agents share the same memory space and actually collaborate instead of duplicating work. Two env vars. Under 5 minutes to connect. Works with Claude Code, Cursor, CrewAI, LangGraph, and AutoGen. What are you using for agent memory across your team right now? Would love to know 👇

GitHub for AI Agent Memory - Shared, versioned memory for your agents and your team

Replies