API

The memo.st API stores durable facts that agents can retrieve across sessions. The API worker lives in apps/api, runs on Cloudflare Workers, and exposes JSON endpoints under /v1/memories.

Memory model

Every memory is scoped by the runtime boundary used by the API:

• ownerId is the Clerk user or organization boundary.
• agentId identifies which agent owns or uses the memory stream.
• pid separates long-running processes, projects, or workspaces.
• tid narrows scope to a thread or conversation.
• subjectId identifies the person, account, project, or entity the fact is about.

The API also keeps namespace as a derived compatibility field, but pid and tid are the primary retrieval scope.

Base routes

GET  /health
GET  /v1/memories
POST /v1/memories
GET  /v1/memories/:id
POST /v1/memories/search

First memory

{
  "agentId": "support-agent",
  "subjectId": "customer_acme",
  "pid": "support",
  "content": "Acme prefers Slack escalation for P0 incidents."
}

The API generates embeddings automatically and can extract KG triples from the same text when enabled.