This first request shows what Augenta does with your data. You send a message to Consolidate; Augenta extracts claims (statements worth remembering) and groups related claims into engrams (consolidated memory with a summary). Nothing is stored: this request only processes what you send, so it needs no Workspace.
When you are ready to give your agent memory that persists, continue with Add memory to your agent.
Create a key
Sign in to the Console and choose Create API key, or open API Keys → New API Key. A key with no Workspace selected can make this request. The secret is shown once: store it in your secret manager and keep it on your server, never in client code.
Send a message
Replace YOUR_API_KEY and run one of these. messages takes chat messages in the OpenAI or
Anthropic format, as you already send them to your model. session is your name for the
conversation; Augenta does not store it here.
curl --fail-with-body -sS "https://api.augenta.ai/v1/consolidate" \
-H "Authorization: AugentaKey YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"session": "hello-world",
"messages": [
{
"role": "user",
"content": "Production releases require owner approval. Record a rollback image digest before deployment."
}
]
}'
Read the response
The response can take several seconds while a model reads your message. The parts to look at:
| Field | What it holds |
|---|---|
extractedClaims.data[].text | Each statement Augenta extracted from your message. |
engrams.data[].summary | Each group of related claims, summarized. |
engrams.data[].claims.data[].text | The claims that support that summary. |
View an example completed response
{
"experiences": [
{
"id": "exp_222222222222222222222222",
"status": "consolidated",
"claimIds": [
"clm_111111111111111111111111"
],
"object": "experience"
}
],
"engrams": {
"object": "list",
"data": [
{
"object": "engram",
"id": "eng_333333333333333333333333",
"summary": "Production releases need owner approval and a recorded rollback image digest.",
"label": "releases \u00b7 rollback",
"labelSupport": 1,
"memoryClass": "semantic",
"status": "active",
"strength": 1.5,
"heatShare": 1,
"claimCount": 1,
"claimTokenCount": 15,
"bornAt": "2026-10-28T12:00:00Z",
"updatedAt": "2026-10-28T12:00:00Z",
"claims": {
"object": "list",
"data": [
{
"id": "clm_111111111111111111111111",
"text": "Production releases require owner approval and a recorded rollback image digest.",
"experience": "exp_222222222222222222222222",
"at": "2026-10-28T12:00:00Z",
"object": "claim"
}
],
"nextPageUrl": null,
"truncated": false
}
}
],
"nextPageUrl": null,
"truncated": false
},
"timing": {
"firstClaimMs": 12,
"firstEngramMs": 17,
"totalMs": 21
},
"usage": {
"tokensSent": 13,
"tokensSentSource": "estimate:utf8-bytes/3.5"
},
"extractedClaims": {
"object": "list",
"data": [
{
"id": "clm_111111111111111111111111",
"text": "Production releases require owner approval and a recorded rollback image digest.",
"experience": "exp_222222222222222222222222",
"at": "2026-10-28T12:00:00Z",
"object": "claim"
}
],
"nextPageUrl": null,
"truncated": false
},
"extractedClaimsComplete": true,
"events": []
}
The same claim can appear in extractedClaims and inside an engram. IDs in this response
identify objects within it; nothing here can be looked up later. See
the Engram object for every attribute, and Errors if
the request fails.
Next
- Add memory to your agent: recall memory before each model call and add each turn after it, so your agent remembers across conversations.
- Import documents into a Workspace: load existing documents and recall from them.