Ragionex Open-Sources Agent Memory Network: Auto-Learning, Shared, Versioned
Ragionex · reddit · 2026-10-29
A developer has released Ragionex Memory, which lets multiple agents automatically build and maintain a shared knowledge network as they work:
- Auto-learning: agents write documents, design decisions, preferences, and experiment outcomes into the network automatically, with updates, merges, and cleanup of stale content handled automatically
- Context engine: query via natural language through an MCP server; semantically matches even when phrased differently, returns thousands of records in milliseconds, and keeps context lean
- Verbatim storage: no summarizing or rewriting — what goes in is what comes out
- Version control: every edit keeps the old version, recoverable for 30 days; deleted items go to a recycle bin, also held for 30 days
- Team collaboration: each project gets an isolated Vault, with teammates invited and granted read/write/delete permissions
- Privacy: data stored in the EU, AES-256 encrypted, never sent to external AI services or used for training
- Free tier: 10,000 requests per month, unlimited memories, 2 Vaults, and 2 API keys
Related event: Ragionex Open-Sources Agent Memory Network with Auto-Learning and Sharing(2 posts)→
More from coding & agent
- Open-Source Orinyx Launches MCP Tool to Independently Verify Healthcare AI Agent Outputs — Practical_Green5775 · 2026-10-29
- DHH: AI Tools Could Make the Best-vs-Worst Programmer Gap 1,000x — tristanbob · 2026-10-29
- Open-Source Stackchan Desk Robot Chats, Dances and Shows Memes — Full Build Guide Included — alexcovo_eth · 2026-10-29
- Addy Osmani: You're overprompting Claude — one goal plus a check beats prompt bloat — addyosmani · 2026-10-29
- Teach Hermes OPNsense docs with /learn, get an optimized firewall in 15 minutes — Teknium · 2026-10-29
- Interactive ghost story built on Gemma 4 stress-tested with 478 adversarial prompts — Extreme_Ad5709 · 2026-10-29