Seven surfaces, built around a single idea: the AI advisor should already have the full picture before you ask the question.
Visit any company page — LinkedIn, Greenhouse, Lever, Workday, Ashby, or a generic site — and the side panel detects it automatically. One click triggers research. Nothing fires without a click.
The research pipeline runs Serper for parallel web searches across firmographics, leadership, reviews, job listings, and product overview, then Claude Haiku synthesizes everything into structured JSON. Cached indefinitely — re-researching is on-demand via the Research button so you don't burn API calls.
Detectors: LinkedIn profiles, job postings, company pages — plus Greenhouse, Lever, Workday, Ashby, and a domain-name fallback for anything else.
Save any company or job posting and it becomes a single record — no duplicate sync between a "company" store and a "job" store. Attach a job description and Coop scores it 1–10 against your background, target roles, salary floor, and work arrangement.
The score reflects how you actually feel, not just what the JD says. A rating you set (1–5 stars) post-processes the score — a role you're excited about gets a lift; one you're lukewarm on gets a haircut. The breakdown shows strong fits, red flags, and qualification matches with full reasoning behind each.
Score colors: ≥7.5 green · 6.0–7.4 amber · 4.5–5.9 orange · below 4.5 red.
Every saved company moves through a fully customizable Kanban. Stages are user-defined. Drag-drop reordering. Stage timestamps record themselves in a generic map — so if you rename stages or add new ones, history doesn't break.
Each card tracks whose court the ball is in (my court / their court), surfaces last activity from emails and meetings, and shows the current score at a glance. Gmail and Granola attach automatically — you don't manually log contact.
Kanban and grid views. Filterable by stage, score, and action owner. Stat cards at the top are clickable and drill into the underlying list.
Every saved opportunity that hasn't been triaged lands in a personal queue, pre-scored and annotated with the two or three things that actually matter. Pass or apply in one tap.
Open Application auto-binds Coop in the side panel so you always have full context while you're filling out forms. The queue drains as you triage — the rest of the pipeline moves itself.
Scores show strong fits, red flags, and qualification signals. The queue reorders by score so the best opportunities surface first.
Coop auto-binds to whatever entry you're viewing. When you open the chat, he already has your professional profile, the company research, the job description, fit scoring, every email thread with people at that company (via Gmail OAuth), every calendar event, and every Granola meeting transcript.
You stop explaining and start asking. "Given what we talked about with Karri last week, what should I lead with in my application?" And Coop actually knows — not from a prompt you pasted, but from data the extension assembled in the background. You can manually pin Coop to a specific entry with a paperclip button when you want focus across surfaces.
Model fallback chain: GPT-4.1 mini → Claude Haiku → Sonnet → GPT-4.1. UI always surfaces which model actually answered. BYO API keys.
A single user-editable textarea in settings carries every interpretation Coop applies to your data. Salary floors, dealbreakers, draft-vs-evaluate behavior, how aggressive you want scoring to be — all of it lives in plain text you can rewrite.
The code emits neutral facts. The textarea tells Coop how to read them. If you want him sharper, rewrite the textarea. If you want him to never refuse a draft request, write that. No code change, no deploy.
This design comes directly from a painful lesson: hardcoding opinions into prompts means the tool argues with the person using it. Ripping that out and handing interpretation to a user-editable config is one of the best decisions in the whole build.
Coop is a Chrome extension with no server-side component. All pipeline data lives in chrome.storage.local. API keys are user-supplied via an Integrations page — never hardcoded, never leaving the machine.
Gmail access uses Chrome's built-in OAuth (chrome.identity). Granola connects via REST API key. No data passes through a Coop.ai server because there is no Coop.ai server.
Manifest V3 service worker handles all side effects. Every UI surface talks to it through chrome.runtime.sendMessage. Research cached 24h. No API call ever fires without an explicit user action.
Coop is in private beta. Get on the list to be in the next group.
Get beta access →