The AI agent memory API.

01Quickstart

Ingest a turn, recall what matters, and build a token-budgeted context block — three endpoints, any language over HTTP.

curl https://agentprizm.com/api/v1/agent/recall \
  -H "Authorization: Bearer ap_..." \
  -H "Content-Type: application/json" \
  -d '{"query":"renewal blockers","searchMode":"hybrid"}'

02Memory scopes

Scope memory per user, agent, organization, workspace, and environment. Isolation is the default; cross-scope queries are explicit and logged.

03Governance primitives

• Confidence score on every memory
• Fact-validity windows (valid_from / valid_to)
• Contradiction handling and decay
• Audit receipt on every recall
• Right-to-forget in one call

04Do not rebuild the memory stack

A DIY vector database gives you retrieval, then leaves you to build extraction, confidence, stale-fact handling, contradictions, deletion, audit receipts, and permissions. AgentPrizm ships all of it behind one API — see how it compares.