Skip to main content
The AtomicMemory Core REST API is a JSON-over-HTTP interface to the memory engine that backs your AI agents. Every operation — storing a memory, searching for context, or packaging facts into a prompt-ready string — goes through a versioned endpoint under /v1. You can run the server locally via Docker or point your requests at a hosted deployment; the API surface is identical in both cases.

​
Base URL

The default local base URL is http://127.0.0.1:17350. When you deploy AtomicMemory to a hosted environment, replace this with your instance’s public URL. You can override the client-side default by setting the ATOMICMEMORY_API_URL environment variable in any application that consumes the API.
EnvironmentBase URL
Local Docker (default)http://127.0.0.1:17350
Self-hosted / cloudhttps://your-core-url

​
Request Format

All POST requests must include a Content-Type: application/json header and a JSON body. Every endpoint under /v1/* requires an Authorization: Bearer <api-key> header. The /health endpoint is the single exception — it requires no authentication and is suitable for load-balancer health checks. 
curl -s http://127.0.0.1:17350/health
A successful health check returns: 
{ "status": "ok" }

​
Response Format

All responses are JSON. Successful responses carry an HTTP 2xx status code. Error responses include a machine-readable error field and a human-readable message field alongside the appropriate 4xx or 5xx status code. 
{
  "error": "UNAUTHORIZED",
  "message": "Missing or invalid Authorization header."
}

​
Endpoint Groups

GroupEndpointsDescription
HealthGET /healthHealth check — no authentication required
IngestPOST /v1/memories/ingest
POST /v1/memories/ingest/quick		Store memories with full extraction or fast dedup
SearchPOST /v1/memories/search
POST /v1/memories/search/fast		Retrieve memories via hybrid or vector-only search
MemoriesGET /v1/memories/list
GET /v1/memories/:id
DELETE /v1/memories/:id		List, fetch, and remove individual memories
ConsolidatePOST /v1/memories/consolidateCompress and package context for agent prompts

Authentication

Learn how to authenticate requests using bearer tokens and manage API keys.

Ingest

Store memories from conversations using full extraction or fast dedup.

Search

Retrieve relevant memories using hybrid or vector-only semantic search.

Memories

List, fetch, and soft-delete individual memory records by scope or ID.

Consolidate

Compress and package memories into a token-budgeted context string for model prompts.