remember

Record something that happened in this project — a decision, a fix, a fact. Your hosted Mneva brain keeps it across every session and every tool.

Signature

remember({ text: string }) → { id: number, text: string, semantic: boolean }

Response:

• id — the new memory's id (per-tenant, monotonic).
• text — echoed back.
• semantic — true if the embedding was written (Ollama up), false if keyword-only fallback.

Example

MCP (from your agent):

> remember the rate limiter middleware lives in lib/rate.ts and is shared by every route group

REST equivalent:

curl -X POST https://mneva.dev/v1/remember \
  -H "x-mneva-key: $MNEVA_KEY" \
  -H "content-type: application/json" \
  -d '{"text":"the rate limiter middleware lives in lib/rate.ts and is shared by every route group"}'

Response:

{
  "id": 47,
  "text": "the rate limiter middleware lives in lib/rate.ts and is shared by every route group",
  "semantic": true
}

Notes

• Append-only. Memories never get edited or deleted. If a fact changes, write a new memory; the old one stays as a historical record.
• For things that can change with evidence, use believe. Beliefs are revisable; memories are not.
• For danger/fragility, use flag. That writes a memory with elevated caution, which assess reads.

See also

• recall — semantic search across memories
• flag — record a memory carrying caution weight
• Memory — the concept page