What your agents shipped

A scorecard of your coding agents: what they asked to measure, what won, what hurt, and what waits for you. On the Agents page, and by email every Monday.

Needlr keeps a scorecard of what your coding agents asked to measure, and what came of it. You see it on the Agents page. Owners and Editors also get it by email every Monday.

Think of a sales report

A sales team doesn't just count calls. Once a week, someone adds up the deals: won, lost, still open. That's how you know if the team is worth it.

The scorecard does that for your agents. Not how much code they wrote, but what it did to your KPIs.

Where you see it

• The top of the Agents page. One line: “Last 30 days: your agents asked about 14 changes. 9 got a verdict: 3 proven wins, 2 likely, 2 hurt. 5 still measuring.” It counts impact checks your agents asked for in the last 30 days, and appears once they've asked about at least one. Everyone on the team sees it.
• Each agent's card. How many impact checks it asked for, how many are measuring, how many have a verdict, and when it last called. These count all time, not just 30 days.
• The Monday email. The same scorecard for the past week, one line per agent.
• The Activity page. Every step each agent took, under its name.
• Each impact check. Its check page says Requested by and the agent's name.

What each number means

Asked about

Impact checks your agents asked for in that period.

Got a verdict

Of those, the ones with a final result, including split tests that stopped because they hurt.

Proven win

An agent's split test that improved the KPI and cleared its bar. A win marked Not confirmed isn't counted (why).

Likely win

A before vs after impact check where the KPI went up, or an adoption impact check that reached its target. A strong signal, not proof.

Hurt

Got worse, plus split tests that stopped themselves because they hurt.

Still measuring

Running, no verdict yet.

The Monday email

• Sent on Mondays, from 08:00 UTC.
• To the account's Owners and Editors who left it on.
• Only when at least one agent is connected and not paused, and your agents asked about something in the past 7 days. A quiet week means no email.
• The subject sums it up: “Your agents this week: 3 proven wins, 1 hurt”.
• Inside: the totals, one line per agent, how many things wait for a person (tests to start, KPIs to create or verdicts to decide), and an Open Agents button.

The email counts impact checks asked for in the past 7 days. A split test needs at least 7 full days, so its verdict usually comes later. Look for it on the Agents page, which covers 30 days.

Turn the email off

1. 1Open Team in the menu on the left.
2. 2Under Your emails, untick Agents weekly email.
3. 3Done. It only changes your own emails in this account. Everyone else keeps theirs, and you can tick it again any time.

How to read it

• Look at what hurt first. Those changes cost you. Make sure they're rolled back.
• Proven beats likely. For a change that matters, a likely win is a reason to run a split test, not to celebrate.
• Few verdicts? Your agents may be asking about changes too small to measure (why).
• Things waiting? Clear them in Waiting for you.

Common mistakes

• Counting “asked about” as value. That's effort. Wins are value.
• Judging an agent on one week. Split tests take a week or more. Look at the 30-day line too.