Work Your Cogny Tickets From Claude, ChatGPT or Cursor — Cogny Keeps the Record

A lot of teams we talk to already have an AI they use all day. Marketing lives in ChatGPT, the developer lives in Cursor, the founder lives in Claude. Asking them to open one more app every morning is a hard sell, and a fair one.

So we stopped asking. As of today, everything Cogny decides needs doing can be done from the AI your team already uses, and Cogny still keeps the record: what was planned, who did it, what changed, and what it did to the numbers.

How it works

Cogny connects to your stack (ad accounts, analytics, Search Console, your CMS, your email and CRM tools) and turns what it finds into tickets with an expected outcome. That part is unchanged.

What's new is where the work happens:

1. The ticket goes to the right person. In Settings → Access & privacy, the workspace owner marks what each member owns: paid media, SEO, content and social, email and CRM, analytics and tracking, website, sales. A new ticket is assigned to the one member whose area fits it and who has access to the tools it needs. If several people fit, it waits for someone to pick it.
2. It reaches them in their own AI. The assignee gets an email with a ready prompt and links that open Claude or ChatGPT with it filled in. The next time they open their AI, Cogny's connector tells it there are tickets waiting, so it brings them up.
3. They do it there. start_ticket gives their AI a brief: the outcome to reach, the tools to use, and the rules (ask before spending money, publishing, sending or deleting). The changes go through Cogny's connected tools, so each one is logged against the ticket.
4. They hand it back. complete_ticket records exactly what changed. Cogny measures the result three days later, writes it on the ticket, and emails the person who did the work: it worked, it partly worked, or it didn't move the number.

The ticket shows who did it and from which AI ("Done by anna@… from Claude"), and an owner can see who on the team works this way in Settings → Access & privacy → Own AI.

Nothing falls through the cracks

People don't always start from a ticket. Sometimes you just ask your AI to fix something. Cogny covers that too:

• On Free, Solo, Cogny AI and Cogny Pro, anything your AI ships or changes through Cogny becomes a ticket on its own and goes straight to measurement: a new campaign, ad, post, page or tag publish, and also a budget change, a bid change, a paused campaign, new keywords or new targeting. Several changes from one task land on one ticket. Anything else, your AI offers to register when it's done.
• On Cloud, changes made while a ticket is open belong to that ticket. Changes made outside one show up in list_unregistered_changes (the last 14 days), and your AI can attach them to a ticket so they get measured.

Ask your AI "what did my changes do?" (the get_my_impact tool) and it answers from Cogny's record for the last 90 days.

Loops, from your AI too

Solo, Cogny AI and Cogny Pro can now create, change, pause and test-run Loops from their own AI as well: a task that runs on a schedule with the tools you approve, up to five at a time, at most hourly. Each run is a ticket that gets measured. On Free, loops are part of Solo.

What each plan gets

The connector works with Claude (web, desktop and Claude Code), ChatGPT, Cursor, Codex and other MCP clients. What every tool does is listed on the integration tool pages, and the plan details are in the buying guide.

Why we built it this way

You can wire your own AI to an ad account without us. What you don't get that way is the part around the change: knowing what's worth doing, giving it to the person who owns that area, and finding out afterwards whether it worked. That's the part Cogny keeps, wherever the work itself happens.

Connect your AI in Settings → Connect AI Clients, or start free.