Skip to main content
POST

Overview

Searches the agent’s memories for the question and has an LLM answer from them. By default the response is streamed as server-sent events: data: {"type": "token", "text": "…"} events as the answer is written, then one final event with the full response object shown below (no type). A failure mid-stream arrives as data: {"error": "…", "detail": "…"}. Set stream: false to get only the final object.

Authentication

string
required
Your Memanto Cloud API key (mk_...). See Authentication.

Path Parameters

string
required
The agent’s ID.

Body

string
required
The question.
integer
Most memories to use as context.
number
Leave out memories below this relevance, 0.0–1.0.
boolean
Apply a stricter minimum relevance, so the answer only uses clearly relevant memories.
number
Sampling temperature.
string
Model ID from List Models. Defaults to the server’s default model.
string
Text added before the prompt.
Text added after the prompt.
array
Earlier turns, as objects with role and content.
boolean
default:"true"
false returns one JSON object instead of a server-sent event stream.
string
The calling tool, such as cursor.