An Issue Tracker Where Your AI Agents Get Assigned Tickets Like Everyone Else
You've probably tried wiring an AI agent into your workflow by now. And you've probably noticed the same awkward gap: the agent lives in a terminal or a chat window, while your actual work lives in a board somewhere else. Two systems, no shared context, and you playing courier between them.
It's a Plan is a self-hosted issue tracker that closes that gap by treating AI agents as first-class teammates. They get a role, permissions, and an assignee slot on the same board as your human contributors.
What It Does
At its core, It's a Plan is a full project management and issue tracking tool. You get projects, Kanban, table, timeline, and calendar views, cycles, custom fields, dashboards, docs, and notes boards. If you never touch the AI features, it still works as a conventional tracker—the README is explicit about that.
The distinguishing part is how agents fit in. An agent is a project member. It can be internal (running on any model you choose) or external (running on your own coding CLI). Runs kick off in three ways: when someone @mentions the agent, when an issue gets assigned to it, or on a schedule you define.
Everything is exposed programmatically. There's a REST API with OpenAPI specs, an MCP server, and webhooks. Pull request integration covers five forges, and authentication supports passkeys and Google sign-in. You self-host it on your own server, with your own database and your own API keys.
Why It's Cool
Agents are members, not a sidebar. Most "AI in your project tool" implementations are bolted-on chat panels. Here, the agent occupies the same conceptual space as a person—same assignment mechanics, same permission model, same board. That's a meaningfully different design decision, and it means you don't have to mentally translate between "the agent's world" and "the team's world."
Scheduled runs are the sleeper feature. Mention-triggered and assignment-triggered agents are useful, but scheduled runs open up a different category of work: recurring triage, status sweeps, periodic cleanup. It's the kind of thing you'd otherwise build as a cron job with duct tape and a prayer.
Everything is scriptable. REST API with OpenAPI, MCP server, webhooks. If you want to build your own agent harness, trigger runs from CI, or pipe board events into some other system, the doors are open. No walled garden.
No per-seat fees, no lock-in. You run it on your infrastructure with your credentials. For teams that have gotten tired of watching SaaS seats multiply, that's the whole pitch in one line.
It's honest about its state. The README says plainly: expect breaking changes before the first stable release. That's a refreshing thing to see stated up front rather than buried in a changelog.
It's not just for engineering. The README lists software development, marketing, design, support, sales, operations, and research as fits. The agent-as-teammate model generalizes—any recurring workflow that produces issues could plausibly be handed to an agent.
The tradeoff worth noting: self-hosting means you own the operational burden. Database, updates, backups, the whole thing. That's the price of the control you get.
How to Try It
The fastest path is a one-click deploy, and the project offers several:
- Deploy on Railway via the button in the README.
- Deploy on Coolify using the guide at
docs/coolify.md. - Self-host with Docker following
docs/self-hosting.md.
If you'd rather poke around first, the project site is at itsaplan.dev, and the source lives at github.com/croffasia/itsaplan.
Once you're running, the natural first experiment is to create an agent, give it a role, and assign it a real issue. Then try a scheduled run and see what it does with a recurring task. The README also points to a Discussions tab if you hit questions, and CONTRIBUTING.md if you want to send patches.
If it looks useful, starring the repo is the lowest-cost way to help—it's what makes the project findable for other people running teams.
Final Thoughts
It's a Plan is best suited to teams who already self-host some of their tooling, have opinions about where their data lives, and are actively experimenting with agents in their workflow. If you're in that group, the agent-as-teammate model is worth an afternoon of your time to evaluate—especially if scheduled runs map onto something you're currently doing manually.
It's under active development, so don't migrate your company's entire process onto it tomorrow. But as a place to explore what agent-in-the-loop project management actually feels like when the agent isn't a novelty feature, it's a genuinely interesting piece of software.
Follow @githubprojects for more developer tools and open source projects.