MCP Tools
extract_memories
Extract entities and facts from text using LLM analysis
Extract entities, facts, and relationships from text using Claude. Requires ENABLE_MEMORY_EXTRACTION=true and ANTHROPIC_API_KEY.
Parameters
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
text | string | REQUIRED | - | Text to extract from (10-100000 chars) |
source | enum | OPTIONAL | manual | Source: conversation, note, document, manual |
storeResults | boolean | OPTIONAL | true | Store extracted memories in database |
Example Request
{
"text": "Ada prefers TypeScript and uses Supabase for his projects. He works at Acme Corp.",
"source": "conversation",
"storeResults": false
}Response
Returns extracted entities with their types, observations, and relationships.
{
"entities": [
{ "name": "Ada", "type": "person", "observations": ["prefers TypeScript", "uses Supabase", "works at Acme Corp"] }
],
"relations": [
{ "from": "Ada", "relation": "works_at", "to": "Acme Corp" }
],
"stored": false
}Set storeResults: true to persist extracted memories to the database.
Errors
| Error | Cause | Fix |
|---|---|---|
| Memory extraction not enabled | ENABLE_MEMORY_EXTRACTION is false | Set ENABLE_MEMORY_EXTRACTION=true |
| Missing API key | No ANTHROPIC_API_KEY configured | Set ANTHROPIC_API_KEY in .env |
Related Tools
remember_fact- Manually store individual factsquery_memory- Search extracted memoriesbuild_knowledge- Batch store facts and relations