Give your AI agents persistent memory. Store, retrieve, and recall context across every session — semantically, at the edge, in milliseconds.
Send any text. We embed it automatically.
store({ userId, agentId, content })Search by meaning, ranked by relevance.
retrieve({ userId, query })Get context ready for a system prompt.
context({ userId, format:"anthropic" })
Add persistent memory to any Claude Managed Agent in one line of JSON.
See the integration guide →{
"mcp_servers": [{
"type": "url",
"name": "agentmemo",
"url": "https://agentmemo.dev/mcp"
}]
}The only agent memory API with OWASP ASI06 protection built in.
Every write scored; low-trust writes are flagged or blocked.
Every operation logged — who, what, when.
Malicious and contradictory writes detected.
Meaning, not keywords.
Ready for LLM injection.
Memories that self-clean.
Organized memory spaces.
Works in Claude and Cursor.
Semantic, episodic, procedural, working, emotional.
# store, then retrieve curl -X POST https://agentmemo.dev/memory/store \ -H "Authorization: Bearer am_sk_..." \ -d '{"user_id":"u1","agent_id":"a1","content":"Prefers dark mode"}' curl "https://agentmemo.dev/memory/retrieve?user_id=u1&q=preferences" \ -H "Authorization: Bearer am_sk_..."
| Feature | AgentMemo | Mem0 | Zep |
|---|---|---|---|
| Edge deployment | ✅ | ❌ | ❌ |
| MCP native | ✅ | ❌ | ❌ |
| auth.md support | ✅ | ❌ | ❌ |
| OWASP ASI06 protection | ✅ | ❌ | ❌ |
| Full audit trail | ✅ | ❌ | ❌ |
| Free tier | ✅ unlimited (beta) | ✅ limited | ❌ |
| Graph memory | ✅ basic | paid | ✅ |
| Temporal knowledge graph | roadmap | ❌ | ✅ |
| Open source | ✅ | ✅ | partial |
Plans below are planned post-beta pricing.