Filter and page facts and stacks over the API and MCP

Facts filter by type and subject, stacks filter by title, and fact lists page with an offset, across REST, MCP and the assistant's recall, for builders working with NOAN's facts.

Until now, reading facts and stacks over the API or MCP meant pulling everything back and sorting it yourself. There was no way to ask for only the facts about a given subject, no way to narrow stacks by name, and no stable way to page through a long list.

What's new

• Filter facts by type and subject. GET /v1/facts, the MCP list_facts tool and the assistant's fact recall now accept a type (primary type) and a subject (matched regardless of case), so you can pull just the pricing facts, just the people, just one subject.
• Page through facts. list_facts now takes an offset, so you can walk a large project a page at a time.
• Filter stacks by title. list_stacks now filters by title.
• More in the stack response. POST /stacks now returns each block's description and whether it is managed.
• Notes carry their externalId. Note reads now include externalId.
• Stable ordering. Contact, note and asset lists break sort ties by id, so paging returns each record once.

Why it matters

If you are building an agent that acts on a project's facts, this is the difference between reading the whole project on every run and asking for exactly the facts you need. Combined with the classifications NOAN now assigns, an agent can request the pricing facts for one product and leave everything else alone.

Where to find it

The filters and fields are documented in the API docs at api.getnoan.com, and the same filters are available through the MCP tools.

Good to know: the type filter matches the primary type only, and subject is case-insensitive.