Turn files, messages, docs, and voice into Unified Memory Objects your agents recall through the REST API + MCP. Capture once, recall many, and show a receipt for every answer.
# capture a memory → get a UMO back curl https://api.memoryintelligence.io/v1/process \ -H "Authorization: Bearer mi_sk_beta_…" \ -H "Content-Type: application/json" \ -d '{"content": "Shipping moved to Tuesdays."}' # → { "status": "success", "data": { "umo_id": "8f3f…", "quality_score": 0.82 } }
A unit of meaning with a receipt.
A UMO is your content, turned into meaning your agent can trust. Give capture a note and you get back a structured, addressable memory it can recall, cite, and prove.
id across sessions and apps. Explain it, forget it, export it.// a Unified Memory Object { "id": "019f3881-…", "summary": "Dana proposed moving the launch to Friday.", "entities": [ { "name": "Dana", "type": "person" } ], "relationships": [ "Dana → proposed → moving the launch" ], "topics": ["launch", "planning"], "tone": "positive", "quality": 0.82, "receipt": { "verified": true } }
Illustrative shape. You address it by id; verify() checks the receipt. The API Reference has the exact fields.
Three doors into the API. Most builders start with the API reference.
The core methods. Click one to watch it run in a terminal.
Typed clients for Python and JavaScript, live now. The REST API and MCP work too.
Point Claude, ChatGPT, or Cursor at the whole API and let it write correct integration code on the first try. MemoryIntelligence ships LLM-native docs.