feed

Extract facts from text or a chat, then store them. REST only — there is no MCP feed tool.

POST/v1/feed

Send either content or messages. If both are present, content is used.

Field	Type	Required	Description
`content`	string	no	Raw text. Max 5,000 bytes.
`messages`	Message[]	no	Chat turns. Max 20. Each message has `role` (`user`, `assistant`, or `system`) and `content`.
`source`	string	no	Label for the extracted facts. Max 255.
`collection`	string	no	Collection code. Omit for the connection default. `""` forces the default bucket.
`reason`	boolean	no	Default `false`: one-shot extraction, then store. `true`: an agent recalls what is already in the graph, keeps durable facts, namespaces `type:id`, normalizes abbreviations, and skips duplicates. Nested remember/recall bill their own credits on top of the reason base.

Request — text

{
  "content": "Maya works at Acme and writes TypeScript.",
  "source": "onboarding-conversation"
}

Request — messages

{
  "messages": [
    { "role": "user", "content": "I work at Acme as a backend engineer" },
    { "role": "assistant", "content": "What stack?" },
    { "role": "user", "content": "Mostly Go and PostgreSQL" }
  ],
  "source": "onboarding_chat"
}

Request — reason

{
  "content": "Maya works at Acme and writes TypeScript.",
  "source": "onboarding-conversation",
  "reason": true
}

SDKs:

await memory.feedWithOptions({
  content: "Maya works at Acme and writes TypeScript.",
  source: "onboarding-conversation",
  reason: true,
});

memory.FeedWithOptions(ctx, gomind.FeedRequest{
  Content: "Maya works at Acme and writes TypeScript.",
  Source:  "onboarding-conversation",
  Reason:  true,
})

$memory->feed(
  'Maya works at Acme and writes TypeScript.',
  'onboarding-conversation',
  null,
  true
);

Response

{
  "status": "completed",
  "facts_extracted": 3,
  "facts_created": 3,
  "facts": []
}

If you already have triples, skip feed and call remember_many.