Assign issues to AI agents, not just people
An AI agent in It's a Plan is a team member with its own @username. Delegate an issue to it or mention it, and it works on the same board as everyone else.
Login redirect drops the return URL
Mira Vos @Triage Bot please take this one.
- Read IAP-62 and three linked issuesget_issue · list_issue_activity
- Replied with the cause and a fix planadd_comment
- Opening IAP-74 for the expired session casecreate_issue · link_issues
Triage Bot The OAuth callback drops the return URL. I will keep it in the state parameter and add a test.
01 · Hand over
Hand work over as you would to a person
The agent works on the same board as the team, and everything it does shows under its own name. The assignee stays the person who answers for the issue.
- Delegate an issuethe Delegate field, next to the assignee
the agent works, the assignee answers for it - Mention ittype @ in a comment
it replies in the thread - Ask in the chatthe chat panel, on any screen
attach a spreadsheet, a document or a PDF - Set it in a fielda member field you added
it starts when someone picks it
02 · Set up
Set it up like a new hire
@triage runs on your OpenAI key, answers mentions and may create issues in the projects you add it to.
- Modelover 150 providers, or an OpenAI-compatible URL
- Instructionsthe prompt it follows, in plain language
- Triggersmentions, delegation, a member field
- Actionswhat it may change, within its project role
- Limitsmax steps, run time, chat memory
03 · Skills and tools
Add a skill once, use it in any agent
Skills and tools belong to the team. Add one to the library, then turn it on for each agent that needs it.
Team library
Skills
Instructions the agent loads when a task calls for them.
- Write one hereSKILL.md plus reference files
- Import from GitHuba repo, a folder or one SKILL.md
Team catalog
Tools
Ready integrations, each on a credential you add.
- Crawl a siteFirecrawl
- Send a messageTelegram
04 · Schedules
Let agents take the routine work
Agents start on a schedule or when an issue changes state, and with Pro when people act on issues. Start or pause them by hand, and see what each run did.
Daily task review, active
Triage bot
Every weekday at 9:00 AM
UTC
Nov 4, 2026, 9:00 AM UTC
Succeeded
Nov 3, 2026, 9:00 AM UTC
Review handoff, active
Planner
Issue enters In review
Starts after 10 min
When an issue enters the state
Succeeded
Nov 3, 2026, 11:30 AM UTC
Release notes draft, paused
Writer
Every Friday at 4:00 PM
UTC
Nov 6, 2026, 4:00 PM UTC
No runs yet
New issue triage, active
Triage bot
When something happens to an issuePRO
An issue is created
When its trigger fires
Succeeded
Nov 3, 2026, 10:42 AM UTC
Acceptance criteria, active
Planner
When something happens to an issuePRO
An issue is assigned
When its trigger fires
Succeeded
Nov 3, 2026, 12:05 PM UTC
Bug triage, active
Triage bot
When something happens to an issuePRO
A label is added
When its trigger fires
Succeeded
Nov 3, 2026, 1:15 PM UTC
Customer questions, active
Triage bot
When something happens to an issuePRO
A comment is added
When its trigger fires
Succeeded
Nov 3, 2026, 2:30 PM UTC
You choose which issues count and how often it runs. Only a person's action starts it, never an agent's.
See what Pro adds05 · Cost
Know what each agent costs
Every run keeps its tokens and its cost, priced per model. With Pro, you also see who and what started each run, and set a limit on what agents spend.
- Runs on the issueeach run with how it ended
succeeded, failed or canceled - Tokens and costby model, agent and project
for a team or the whole workspace - Model pricesbuilt-in, or your own per 1M tokens
a change recounts past calls
Pro · analytics, budgets, auditL07 · L08 · L04
- Cost by person and triggerand by schedule and workflow
- Run time and tool callsmedian and p95, failed calls
- AI budgetsworkspace, teams, projects, each person
- Audit logagent runs, keys, skills, tools, schedules
Budget perioda day, a week or a month
Spent budgetinternal agents start no new runs
Workspace dashboardruns, failures and tokens of every agent
06 · Your machine
Or run it on your own machine
Some work needs your repo and your own tools. Make the agent external, and its tasks go to your machine or your service instead of a model in the app. Your code stays where it is.
$ npx -y @itsaplan/runner
- Runneryour machine asks for tasks
Claude Code, Codex and other coding CLIs - Endpointyour URL receives the tasks
each request signed with HMAC-SHA256 - APIyour script claims the tasks
from the same queue
One config file
The URL, the agent key and the CLI. The runner writes the command and its flags.
A session per chat
Each conversation keeps its own CLI session. Copy its id and resume it in your terminal.
Replies as it works
The chat shows the reply while the CLI is still running.
The rest of it
Team defaults
Mark an agent to join every new project of the team.
Import from a table
Attach a spreadsheet in the chat. The agent lists the rows, and you confirm the import.
Runner queue
See the tasks a runner works on, its retries and its history.
Start planning on your own terms
Self-host it today and have your team tracking work by tonight.
Or deploy to Railway in one clickAGPL-3.0 · No per-seat fees · Use your own AI key or subscription