It's a Plan: The Issue Tracker Where Agents Join the Board

· Tutorials

A self-hosted Linear alternative, seven weeks old at 451 stars, where AI agents get roles, permissions and an assignee slot. MCP built in. Buried Treasure No11.

Last updated: September 6, 2026 · 7-minute read

Seven weeks. That is how long It's a Plan has existed — created July 14, 2026 — and the repository already shows 451 stars, 55 forks, five tagged releases in the last twelve days of August alone, and a feature list that reads like Linear's greatest hits. The twist that earned it a Buried Treasure slot: this is the first self-hosted tracker this series has covered where AI agents are modeled as first-class team members rather than bolted on through an API.

Buried Treasure №11, and a natural sequel to the agency-agents experiment this site covered in August. Every fact below comes from the repository and its README, verified September 6, 2026.

What the repo actually shows

The receipts: AGPL-3.0 license, TypeScript codebase, pushed to on the day of writing, v0.16.0 released August 31 after v0.12, v0.13, v0.14 and v0.15 all landed in the fortnight before it. Nineteen open issues against an actively triaged queue, one-click deploy buttons for Railway, Coolify and plain Docker, and a website at itsaplan.dev. For a seven-week-old project, that is a startling amount of operational maturity.

The positioning is equally direct: an open-source alternative to Linear, Jira, Trello and Plane, self-hosted, where teams and AI agents work side by side. The README's own summary line — run it on your own server, your own database, your own API keys, no per-seat fees — names the exact pain that drives teams off Jira's pricing page.

Agents as teammates, not integrations

Every tracker in 2026 has an AI button. It's a Plan's model is different, and more interesting: an agent gets a role, a permission set, and its own slot in the assignee picker. It takes issues on the same board your people do, gated by the same permissions your contractors have. Work assignment becomes a dropdown decision — this ticket goes to Priya, that one goes to the triage agent.

Underneath, the machine-facing surface is complete rather than decorative: a REST API with OpenAPI specs, webhooks, an MCP server so external assistants can read and change issues through the same protocol, and pull-request integration spanning five forges. That last detail matters — this is a tracker designed for an era when the entity closing your tickets may be running in a terminal on someone's laptop rather than an office.

The full tracker underneath

Strip the agents out and the foundation still has to stand, because no team adopts a permissions system for bots. The base is a full issue tracker: projects, boards, cycles, custom fields, dashboards, gantt and timeline views, sprint planning, bug reporting flows. Passkeys and Google sign-in ship in the box, and the interface localizes into six languages — an unusual depth signal for a project this young.

Deployment is the pragmatic kind. Railway's one-click template carries the whole stack, Coolify gets its own documented path, and the Docker Compose route stays available for people who consider one-click deploys a last resort. None of the paths require reading a wiki first, which puts it ahead of most self-hosted trackers at the same age.

The honest caveats

Three belong in the open. First, the README says it outright: expect breaking changes before the first stable release — v0.16 is a promise of velocity, not stability, and anyone deploying for a real team should pin versions and read release notes like ritual. Second, AGPL-3.0 is a genuine commitment; fine for internal company use, but anyone building a hosted service on top needs to understand the share-alike terms before writing code.

Third, the agent-teammate model is the newest and least battle-tested part of a young project. The concept maps cleanly onto everything this site has learned from multi-agent workflows — agents need scoped permissions and reviewable outputs — and It's a Plan is the first tracker to encode that lesson structurally. Whether the model survives contact with fifty real teams is a question only the issue tracker can answer about itself.

Who it is for, who should skip

Strong fit: small teams already running AI coding agents who want the same actors tracked on the board — the tool finally matches the workflow. Also compelling for Jira refugees on budget grounds, since self-hosting on the same VPS tier as SwiftWave costs less than one seat on most commercial plans. The general open-source SaaS replacement crowd will recognize it instantly from the alternatives guide.

Skip it today if your process depends on Jira-grade admin controls, audit trails, or an ecosystem of plugins — the bones are young and you will feel it. Solo users who just need tasks should also look elsewhere; this tool's reason to exist is the team-plus-agents shape.

Verdict

It's a Plan is the clearest artifact yet of the 2026 thesis that agent infrastructure, not model releases, is where the interesting engineering happens. Seven weeks old, 451 stars, agents on the board, MCP in the box — and a maintainer shipping releases like the deadline is personal. Watch this one monthly; the trajectory says v1.0 lands before the year turns.

Next Buried Treasure: a database GUI endorsed by the PostgreSQL project itself, deploying in one docker command.

---

Not affiliated with the It's a Plan project. Facts verified from the GitHub repository, README and releases on September 6, 2026. More open-source discoveries in the Buried Treasure series.