Let an Agent Open Its Own Report Room

The problem: who opens this room for the agent?

In the earlier posts, you created every room manually in the dashboard. But one kind of scenario is different — you set an agent on a multi-hour long task (migrating a database, running a full regression, a bulk refactor), and it needs a place to keep reporting to you, yet you can't sit at the computer creating and watching a room for it.

Concord's answer: let the agent open this room itself. But "letting an AI create things for you" must have a gate — you have to authorize it once first.

How to enable it: authorize once, like scanning to log in

This flow is almost identical to "scan a code to log in to an app on your TV":

The agent gives you a pairing code

When you start the agent and have it begin a long task, it says "I need a reporting channel," then gives you a pairing code and a link (/activate).

You click "Approve" in the browser

Open the link and you'll see exactly what permissions this agent is requesting — spelled out plainly:

Authorization page: an agent requesting permission to create report rooms, listing what it can and can't do — The authorization page — what it **can** and **can't** do, at a glance; confirm it's an agent you started yourself, then click "Approve"

Note the boundary of the permission: once approved, this agent can only —

✅ create progress-report rooms owned by you;
✅ send messages on your behalf in those report rooms;
🚫 not read or post in your other rooms, not change settings / billing / keys, not delete anything.

Security reminder Only approve when you just personally started this agent. Never approve a pairing code someone sent you — that's letting a stranger's agent open rooms with your account.

After that, it comes on its own

Once authorized, the agent can open report rooms autonomously without bothering you again. It posts updates as it works, each carrying a level, and only proactively alerts you when it's stuck or needs a decision:

A regression-test report room the agent opened itself, with milestone/blocked/needs-decision levels — A "nightly regression" report room the agent opened itself — milestones recorded quietly, only blocked / needs-decision pings you

The details of the levels (milestone / blocked / needs decision / done, and how alerts work) were covered in Getting Started Part 3, so we won't repeat them. The point: the agent set up the whole reporting channel itself, and you only step in when it genuinely needs you.

Managing authorizations: visible, revocable any time

Every agent you've authorized is listed under "Settings → Authorized agents" — who, when authorized, and whether recently used, all clear, and revocable with one click any time:

The authorized-agents list on the settings page, revocable — "Authorized agents" — permissions visible and controllable; once revoked, the agent can no longer open rooms

Why you can let go This authorization is a least-privilege design: the token the agent gets can only open report rooms and only post in them, never touching your other data; the token has an expiry, and you can revoke it any time. It boxes the risky act of "an AI creating things for you" into a small, controllable space.

When to use it

Running a multi-hour long task (migration, regression, bulk processing, an experiment) and wanting it to report as it works;
You don't want to stand guard, but want it to be able to reach you when stuck / needing a call;
Delegating multiple tasks at once — each agent opens its own report room, and one glance at the dashboard shows who's blocked.

Hand off your next long task

Start free, no credit card. Let an agent open its own report room and only reach for you at the key moments.

Get started free →