Skip to main content
Record the user’s work and decisions so future sessions build on them: drafts you wrote (status: drafted — drafts are never sent), pitches sent, replies received, pages published, site fixes, hand-offs, waits, research notes, and declined_offer ids. Never log your own Parse tool calls. Pass brand when the workspace has several brands. Same brand + kind + target appends to the existing item (pass item to target one explicitly; items batches multiple). Free.

Who may call it, and what it costs

Price. Free. Every reply says what the call cost in meta.costCredits. The prices are on Spending, tracking and stopping.

Arguments

one of plan, log, update, close, claim, release
“plan” for a hand-off or follow-up wait (pair with owner and/or due date)
string
what happened or is planned, in your words; pasted email text is fine (up to 4,000 characters)
string
a short title; default the note’s first sentence, cut to 80 characters
string
Work type: pitch, listing, reply, page_written, page_edit, site_fix, declined_offer (omit for plain notes or hand-offs)
string
drafted (for drafts you wrote — never sent), planned, sent, published, replied, declined
string
brd_ id, or the brand’s name or domain; default the workspace’s brand when it has exactly one
string
what it is aimed at: a page URL or pg_ id, a site (domain or site:), a contact’s email or ct_ id, a top_ id
string
the page, as older calls pass it: a pg_ id or URL
string or object
who it went to: an address, a form or author page URL, a ct_ id, or { name, role, email, site }
string
why: a top_ id, or the gap in words
string
why: a qst_ id
string
when to follow up: a date, today, tomorrow, a weekday, next week, in 3 days
string
who does the work, to hand it to someone: a person’s name; default this key’s label on a new item. Who wrote each change is kept on its event
where the work lives: a doc, a PR, a thread
string
Existing itm_… id when recording a reply or follow-up on a prior item
string
nt_ id: file a note kept to file, with what it lacked
boolean or string
true: let status move back or one end change to another, recorded as a correction
string
when it happened: a date, yesterday, a weekday (the last one), last week or 3 days ago; default today. A page that went up yesterday is at yesterday
string
the same as text
one of first_questions, track_weekly, keep_running, top_up, set_up
with kind declined_offer: the offer’s id
list of object
several notes in one call (up to 50), each with the fields above; the call’s own fields apply to each as defaults up to 50
Every argument the tool takes, as GET /v1/tools serves it, every shared object written out; https://agent.parse.gl/openapi.json serves the same schema with the filter and the window named once under components and pointed to by $ref. MCP tools/list leaves out the arguments few calls send, and says the filter and the window once, under search_answers; resolve_ids {"kinds": ["argument"]} lists what it left out, and a call with any of them runs:

Example call

A note as you have it: Parse files it as an item on the page, with the contact, and replies with what it understood. Over REST:
Over MCP, the assistant calls the tool by name with the same arguments:
More calls:
  • Plan several pieces of work at once: to-dos are items in planned: {"move":"plan","brand":"<brd_ id>","items":[{"kind":"pitch","target":"https://example.com/best-tools","due":"2026-11-20"},{"kind":"page_written","topic":"pricing comparison","owner":"Dana"}]}
  • Update an item: a second session or another agent writing the same brand, kind and target updates it too: {"item":"itm_…","status":"replied","text":"The editor wants a quote from the CEO"}
  • The older call, which works as before: a pitch sent to a page: {"kind":"pitched","page":"https://example.com/best-tools","brand":"<brd_ id>","note":"Emailed the editor"}
The reply has data, context, next, page and meta; follow next for the calls that come after. A refused call returns { code, message, fix }: see the error table.