Start on the ticket in the agent's name

  • Desktop app
  • All plans
View as Markdown

What it does

Many people name an agent after the tracker ticket it is created for: ARD-221, PROJ-14 login. With this project setting on, that key becomes the agent's first message at its first launch ("Start work on ticket ARD-221." by default), so the agent starts on the ticket without anyone typing the key it is already named after. It is the missing link for an intake skill or an MCP connector that fetches a Linear, Jira, YouTrack or Shortcut issue once it is told the key: a skill fires on what is said in the conversation, never on the agent's name, and this setting says it. Off by default: a name is just a name.

Where to find it

  • Project settings > Agents & AI > Default agents: the card Start on the ticket in the agent's name, with a First message template under it and its Save button.
  • Nothing to find on the agent itself: the trigger is the Name you give it in the Add agent or Edit agent sheet.

How to use it

  1. Open Project settings > Default agents and switch on Start on the ticket in the agent's name.
  2. Optionally rewrite First message. {key} stands for the ticket key ("{key} stands for the ticket key. Leave empty to keep the default."). A live example under the field shows what an agent named ARD-221 would receive ("An agent named ARD-221 receives: …"). Click Save: a half-typed line never reaches an agent.
  3. Add an agent and put the ticket key in its Name: the whole name (ARD-221) or part of it (ARD-221 login fix, [web-14] header). Launch it: the first message is sent, the agent fetches the ticket with whatever skill or connector it has and starts.

What counts as a key: a team prefix of 2 to 10 letters or digits starting with a letter, a dash, then 1 to 7 digits, anywhere in the name and not glued to other letters or digits (ard-221 is read as ARD-221; A-1, MYARD-221x and a date such as 2026-09-18 are not keys). Only the first key of the name is used. The name scanned is the one you typed, never the role label.

The message is a launch brief like a backlog ticket's: delivered once, at the record's first launch, never replayed when you reopen the agent later. An agent that had already run before you turned the setting on receives nothing, even after a rename; an agent that did receive its brief and is then renamed to another key is briefed on the new one at its next launch. A template without {key} still gets the key, appended on its own line.

Settings

  • ticketNameAutoStart (project, Project settings > Default agents > "Start on the ticket in the agent's name"): the master switch. Absent = off.
  • ticketNamePrompt (project, same card > "First message"): the template, {key} replaced by the ticket key. Absent or blank = "Start work on ticket {key}.". Sent as typed, not translated. Saved with your account, so the project has the same behaviour on every computer.

Agent tools (MCP)

  • settings_get, settings_set with scope: "project" and the keys above: an agent can turn the behaviour on or rewrite the template for the current project.
  • settings_list with scope: "project" lists both keys in the agents group.

Providers

All providers, no difference: the message travels the same path as a backlog ticket brief, so a provider that receives a ticket brief receives this one.

Mobile

No settings screen on the phone, like the other project settings of this family (mission, project skills). The behaviour itself is the same from the phone: an agent created or launched from the mobile companion is spawned by the desktop, so a name carrying a key starts on it when the project setting is on. The Name field of the phone's agent sheet is the one scanned.

Limits

  • Per project, off by default: GPT-5, WEB-3 or API-2 are legitimate agent names that look like keys, so a project that does not name its agents after tickets must leave it off.
  • An agent launched for something else (a backlog ticket, a team step, a QA scenario, a scheduled run, a mockup, an imported configuration) keeps that brief; its name is then a label and is not scanned.
  • Template capped at 2000 characters.
  • One key per agent: the first found in the name.
  • Restart the agent with nothing to resume replays the first message, like any launch brief; a normal reopen does not.
  • The message only names the key: fetching the ticket is the job of your skill or MCP connector, AgentsRoom does not read the tracker for it (see Import Tickets for tickets pulled onto the board).

Common questions

  • My agent is called ARD-221 and does nothing at launch. Check that the project setting is on (Project settings > Default agents), and that the agent had never run before you turned it on: a brief is only delivered at the first launch. Create a new agent with that name to get it.
  • Does it work with a backlog card named after a ticket? No, a card briefs its agent with the ticket itself. This setting reads the name of an agent you created in the room.
  • Which trackers? Any that prints keys as PREFIX-NUMBER: Linear, Jira, YouTrack, Shortcut. The message does not say which tracker; your skill knows where the ticket lives.
  • Can I write my own first message? Yes, the First message field, with {key} where the key goes, saved with the project.
  • Is ard-221 in lowercase accepted? Yes, it is read as ARD-221.

Related

  • Backlog Task Board: tickets that live on the board brief their agent themselves, no naming needed.
  • Import Tickets: bring GitHub, JIRA or Notion issues onto the board instead of naming agents after them.
  • Skills Library: the project skill that fetches the ticket and starts on it is what this first message triggers.
  • Multi-Project Multi-Agent: adding and naming agents in a room.