Persistent semantic memory for AI agents. Write once, retrieve forever.
Free Forever. No credit card.
10,000 requests per month, free.
Your AI stores context from past conversations. Later, it recalls exactly what it needs - project decisions, user preferences, anything you told it before.
Every AI agent has the same problem: it forgets. Loses context between sessions. Hallucinates when conversations get long. We built the memory it never had.
POST any content to /v1/memory/write. Returns immediately with a memory id while our engine processes it in the background.
Processing happens silently in the background. No wait, no blocking - just poll the status endpoint until it's ready.
Your agent asks. We understand what it actually needs. The right memories come back.
Agents need memory that survives across sessions, users, and restarts. Ragionex Memory is the storage and retrieval layer - your agent's model does the reasoning on top.
Every memory lives under a project label your agent passes on each call. Coding assistants switching between repos never mix contexts.
Omit the project on search and Memory looks across everything the user has ever stored - useful for personal preferences that span projects.
Ask "what does the user prefer?" and get back facts stored as "User hates Comic Sans and prefers JetBrains Mono for everything." Semantic search finds the right memory even when the query wording is completely different from how it was stored.
Expose Memory to your agent as an MCP tool. Claude, Cursor, Windsurf, and any MCP-compatible client can read and write memories with one config line.
Label your memories with anything that matters - environment, source, priority, expiry - then run advanced recall queries that surface exactly the right one. Coming on Pro.
Memories belong to the user, not the agent. Switch between Claude, Cursor, or any LLM - or run several in parallel - and they all see the same recall pool.
Free Forever. No credit card. No signup friction. No lock-in - export your data anytime.
Generous limits most developers never reach.