Projektor

Projektor

2 min read Original article ↗

Projektor issue backlog — list view with projects sidebar, issue refs, status, priority, and assignees

Projektor holds issues, boards, sprints and a wiki, and exposes 113 tools across 22 domains over MCP. The agent files the ticket, moves it and writes the page instead of asking you to. Other trackers were built for people and had agent access bolted on later; git-file trackers are agent-native but live inside a single repo. Projektor is agent-native from the schema up, works across projects, and is cheap enough to leave running.

Every action, both surfaces

REST and MCP are two doors into one service layer. An agent is not driving a reduced API built for robots — it calls the code the browser calls. Shipping a feature to one surface and not the other counts as a bug.

You still run the project

Agents do the filing; you do the steering, in a normal web app. Backlog and kanban board, epics, sprint planning, a nested wiki with full-text search, flow metrics. Nothing an agent does is buried in a log.

Built for fleets

An agent registry, file claims and coordination messages let many agents work the same repo in parallel without colliding.

Runs on the edge

One Cloudflare Worker: Hono, D1 for data, KV for cache, R2 for attachments. No servers, no containers. Deploy to your own account in minutes.

Deploy to your own Cloudflare account from a config-only repo. Wrangler provisions D1, KV and R2, applies the migrations and ships. Use the one-click button, run one command, or hand the repo to an agent and say “deploy projektor to my Cloudflare account.”

Your first ten minutes → · Why Projektor looks like this → · Poke around a live instance first → · How and why it was built →

Projektor’s own roadmap is tracked inside Projektor; these docs are generated from the same repo.

Built by Verdient.