AI Memory quickstart

Get an Integration key

Create an Integration in the Conare dashboard. You get a scoped Bearer key (cint_...) — it is shown once and stored only as a SHA-256 digest, so put it in a server-side secret manager. Never ship it to a browser.

Verify the credential

GET /api/v1/status is a side-effect-free smoke test: it validates the key, namespace routing, and the memory-plane hop without creating an end user or consuming quota.

curl https://api.conare.ai/api/v1/status \
  -H "Authorization: Bearer $CONARE_KEY"

Save a memory

Save distilled observations (a sentence or paragraph), not raw events. Saving identical content twice is a no-op (deduped: true).

curl https://api.conare.ai/api/v1/memories \
  -H "Authorization: Bearer $CONARE_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "endUserId": "u_123",
    "content": "User prefers smaller islands, wants to avoid crowds. ASA 104 certified.",
    "containerTag": "profile"
  }'

Recall it

Two retrieval modes, pick per call site:

# Raw hybrid search — sub-second, no LLM. Your agent does the reasoning.
curl https://api.conare.ai/api/v1/search \
  -H "Authorization: Bearer $CONARE_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "endUserId": "u_123", "query": "charter budget boat type", "limit": 10 }'

# Deep recall — LLM-synthesized, citation-grounded answer (~3–6s).
# Inject `answer` into your agent's system prompt at session start.
curl https://api.conare.ai/api/v1/recall \
  -H "Authorization: Bearer $CONARE_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "endUserId": "u_123", "query": "what do we know about this user and their upcoming trip" }'

Every response carries X-Request-Id. Send your own (1–128 chars, alphanumeric first) to correlate a request through the Conare edge and memory plane.

Next: understand Integrations, end users, and containers, or jump to the full API Reference.

Conare Platform TypeScript SDK

⌘I

Get started

Foundations

AI Memory quickstart