Loops Lets Agents Read Campaign Metrics and Work From ChatGPT

Loops added an Email Metrics API on free and paid plans and brought its product into ChatGPT and Codex, the company said in its Oct. 8 changelog. The API returns sends, opens, clicks, unsubscribes, spam reports and hard and soft bounces. Loops said teams can feed them into their own reports or an agent.

Metrics come from three endpoints and count sends, not events

Loops’ API reference lists three read endpoints: one for campaigns, one for transactional emails and one for workflow email nodes. Each returns all-time totals, not a time series. The campaign endpoint returns a 400 for a campaign that has not been sent yet.

The reference says the counters work per send. A send opened five times counts as one open, and a person who received the same email twice counts twice. The transactional endpoint totals every published version of the email. The workflow endpoint works only on send-email nodes and returns zeros for a node that has not sent.

The ChatGPT and Codex connector follows an August Claude connector

Loops said its connector now covers ChatGPT and Codex, after a Claude connector released in August. Users can create a draft, update an email or review workflows from a conversation. Loops’ own sample prompt asks an agent to compare click and open rates across the last five campaigns, flag high bounce rates and suggest three tests.

The release also adds API key history and CLI shortcuts

The Settings page now shows who created each new API key and when. The CLI gained view and get --web to open campaigns, workflows and transactional emails in a browser. Emails created through the API without a group now go into Unsorted. Loops also fixed reversed mobile columns in Outlook for Windows.

The take

We think the metrics API is useful, but only if a team reads its numbers the way Loops defines them. Each count is per send. An open rate is not a count of readings. Loops’ own guide calls open rates a vanity metric. It cites privacy changes such as Apple Mail Privacy Protection. An agent asked to “flag low open rates” will rank campaigns on a number the vendor does not trust.

All-time totals cannot show whether a bounce spike happened last week or last year. A script has to save a snapshot each day and subtract. That is easy to build and easy to forget.

Teams on Loops should point the agent at bounces, spam reports and unsubscribes first, store a daily snapshot of the totals, and treat clicks as the engagement number.