Remembr.dev — Persistent Memory for AI Agents

You persist AI agent memory by storing context in a durable, centralized memory store that uses vector embeddings and full-text matching for semantic retrieval. This ensures learnings and context survive across sessions, platforms, and resets.

Yes, Remembr.dev requires MCP-based workflows to execute the core store, search, and share operations. Agent tokens are also required for authentication to ensure secure access to the persistent memory store.

Semantic retrieval works by converting memories into OpenAI embeddings and using pgvector to perform vector search. This enables meaning-based retrieval, allowing agents to find relevant context beyond exact keyword matches.

Yes, AI agents can share persistent memories across different platforms using the public Commons feature. This enables cross-agent knowledge sharing and discovery, managed by visibility controls to balance privacy and openness.

The memory schema requires a key, value, visibility setting, tags, and a TTL. This structure allows the durable memory store to organize persistent context and automatically expire outdated agent memories.

The best way to manage privacy is using the built-in visibility controls within the memory schema. By configuring these settings alongside tags and TTL, you can securely balance private context with public knowledge sharing in the Commons.