Skip to content

Agent Memory

Agent memory lets an agent remember information across separate conversations. Use these endpoints to read or clear stored memories.

Base URL: https://agent-api.auteryn.ai


List memories — GET /api/agents/{agent_id}/memories

Section titled “List memories — GET /api/agents/{agent_id}/memories”
Terminal window
GET https://agent-api.auteryn.ai/api/agents/{agent_id}/memories
Authorization: Bearer <jwt>
X-Agent-Id: {agent_id}

Returns { "memories": [...], "count": N }: what the agent has stored.

Param Description
type Only this kind: agent_knowledge, user_preference, workflow_pattern, or user_feedback
user_id Only this person’s preferences (owners and admins; a member sees their own)
include_deleted Include deleted memories that can still be restored (default false)

Delete memories — DELETE /api/agents/{agent_id}/memories

Section titled “Delete memories — DELETE /api/agents/{agent_id}/memories”

Deletes the agent’s memories: all of them, or only those matching the type and user_id params. A deleted memory can be restored for 30 days. Deleting an agent’s memories needs Run & edit access to the agent; a member may always clear their own preferences.

{
"success": true,
"deleted": 12,
"message": "Deleted 12 memories"
}

Knowledge retrieval (RAG) — separate endpoint

Section titled “Knowledge retrieval (RAG) — separate endpoint”

Searching an agent’s Knowledge uses a different endpoint. The Knowledge comes from the agent in the X-Agent-Id header:

Terminal window
GET https://agent-api.auteryn.ai/api/retrieve?query=your+search+terms&top_k=5
Authorization: Bearer <jwt>
X-Agent-Id: your-agent-uuid
Param Description
query Search query (required)
top_k Number of chunks to return (default: 5)
{
"query": "your search terms",
"results": [
{ "text": "…", "score": 0.82, "metadata": {} }
]
}

To search one Knowledge by its ID:

Terminal window
GET https://agent-api.auteryn.ai/api/knowledge-bases/{kb_id}/retrieve?query=...

Both retrieval endpoints allow 60 requests a minute and refuse guests, customers and agent tokens. This is not the same as agent long-term memory.