The product

How Coop works

Seven pieces. One idea: the advisor already has the full picture before you ask.

Land on a page. Get the company.

Open a job posting and the side panel names the company before you finish reading the title. LinkedIn, Greenhouse, Lever, Workday, Ashby, or any random careers page.

Detection is free and automatic. Everything else waits. Nothing fires until you ask.

LinkedIn Greenhouse Lever Workday Ashby Any page
Coop's side panel open on a job page with the company detected

One record per company. Scored against what you care about.

The research, the role brief, the people, every email and meeting - one record. No syncing a company store against a jobs store.

Three scores sit on top of it:

  • Match - is this the right role? Scored out of 10 against your rubric.
  • Conversation health - how the talks are actually going, 0 to 100.
  • Offer likelihood - a real shot? A banded percent that only hits 100 with a written offer in hand.
An opportunity record for Mercury Labs - research, role brief, contacts, and all three scores

A pipeline that stays honest

Every live conversation on one board, in the stage it actually occupies. Not the stage you wish it occupied.

New emails and meetings attach themselves to the right card. A freshness sweep keeps the whole board current, and it costs nothing to run. You never log contact by hand.

Convo Started Screening Call HM Call Work Sample Panel Decision Offer
The pipeline board with scored opportunity cards across the active stages

Triage one card at a time.

Everything you saved but haven't decided on waits in the Apply Queue. One card at a time, pre-scored, flagged with the two or three things that matter. Pass or apply in one tap.

Best-scored roles come up first. The queue drains. The pipeline fills.

Try the swipe demo →

The Apply Queue - a fit-scored card waiting on a pass-or-apply call

Three verbs, everywhere.

Every entry gets the same three moves. Learn them once.

Refresh re-reads new signal - meetings, emails, elapsed time - then re-runs the research and the scores.

Draft is the outgoing menu: reply, meeting prep, follow-up, cover letter, tailored resume. All of it lands as "Draft - not sent". Coop never sends.

Ask opens a chat seeded with the full entry. It changes nothing.

Mercury Labs Enterprise AE

Folded in 2 new meetings, rescored - Match 7.8 → 8.2 · now B+

Reply · Meeting prep · Follow-up · Cover letter · Tailored resume

“How’s Mercury Labs really going?” He answers. Nothing changes.

Code carries mechanics. Settings carry opinions.

The code emits neutral data. Your operating principles tell Coop how to read it. Salary floor, dealbreakers, how sharp you want the scoring - plain sentences in one editable textarea.

Change a line and every score, draft, and brief downstream changes with it. No code change, no deploy.

The operating principles editor - plain-sentence opinions that steer scoring and drafting

No backend. No account.

Coop is a Chrome extension plus a small MCP companion on your machine. That's the whole architecture. 78 MCP tools and 22 slash prompts hand your Claude the full picture.

Writes are graded. Most land immediately with a receipt and one-tap undo on the From Claude page. Five classes wait for you: outbound sends, identity, money, deletes, and net-new companies.

Zero API keys - reasoning rides the Claude subscription you already pay for. Coop sends nothing anywhere - only your own Claude talks to Anthropic.

The spec

  • Runs asChrome extension + local MCP companion
  • MCP tools78
  • Slash prompts22
  • API keysZero - rides your Claude subscription
  • WritesGraded auto-apply, receipt + undo
  • Waits for youSends, identity, money, deletes, new companies
  • BackendNone. No account either.

/coop:brief · /coop:cover_letter · /coop:weekly_review · /coop:call_prep

Want it for your own search?

Private beta. The form takes about 60 seconds. I review every application personally.

Get beta access