I wired Coop into Gmail, Google Calendar, Granola, and whatever job page you're on. It reads what's already there and attaches it to the right company. No copy-paste, no manual logging, no data leaving your browser.
By the time you've read the description, Coop has pulled your Gmail threads with that domain, your upcoming calendar events, any Granola transcript that mentions them, and scored the fit against your rubric. It all just shows up.
Coop's default brain is Claude Desktop, plugged in via MCP. No API keys to configure. The rest - Gmail, Calendar, Granola - pull straight into browser-local storage. Every byte stays on your machine.
A content script reads LinkedIn, Greenhouse, Lever, Workday, and Ashby pages directly. Open a job posting and Coop pulls the title, JD, comp band, and requirements in the background. Nothing to click, nothing to paste.
No key needed to start. Ask Claude Desktop to research a company and it calls Coop's MCP, pulling firmographics, leadership, reviews, and recent news on demand. Add a Serper key and the same research fires automatically when you open a new opportunity. Cached until you force-refresh from the side panel.
I wired Gmail so it reads threads where the sender or recipient domain matches the company. It surfaces who you've corresponded with, flags rejection language, and drives the "last activity" and "Action on" signals on each kanban card.
Reads events where attendees share the company's domain. Drives "next meeting," the upcoming-events sidebar, and the relationship timeline. Past meetings backfill your history with that company automatically - no manual entry.
Granola transcripts attach to companies automatically. Coop searches by company name and contact name, then pulls the exact paragraph that mentions them - verbatim, their own words. Indexed once on first connect, then queried per opportunity.
Coop registers as an MCP connector in Claude Desktop. Ask Claude to score an opportunity, draft outreach, or debrief a call, and it calls Coop in the background to pull the context. No API keys to get started. It runs through the same Claude subscription you already have.
Add an Anthropic key if you want background work that fires without opening Claude Desktop: fresh-data sweeps, auto-scoring your queue, scheduled drafts. Same Claude under the hood, just the auto-fire path instead of the on-demand one.
Prefer GPT for the auto-fire path? Bring an OpenAI key and Coop routes background work through it instead. Configure both providers and Coop cycles between them on rate limits. The UI shows you which model answered.