projektor MCP
A self-hosted issue tracker and wiki on Cloudflare with MCP: 126 tools, issue leases and file claims for several agents
- 66 ★ on the repository
- Last commit
Broad access
Why this level
- MCP exposes 126 tools that create, change and delete issues, documents and file claims in the workspace
- The workspace token and Worker address give full access to tracker data, and without Cloudflare Access the interface is open to anyone
Install
- Works in
Manual install
claude mcp add --transport http projektor "https://<host>/mcp/<workspace-id>"Deploy an instance on Cloudflare first from the guide; headless agents need a workspace token.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- MCP exposes 126 tools that create, change and delete issues, documents and file claims in the workspace
- The workspace token and Worker address give full access to tracker data, and without Cloudflare Access the interface is open to anyone
This is third-party code. Review the repository files before installing.
What it does
projektor is an issue tracker and wiki that an AI agent can run as well as you do. Every feature is exposed over MCP: the README counts 126 tools across 24 domains, so the agent files, moves and documents the work itself. The app runs as one Cloudflare Worker in your own account on `Hono`, `D1`, `KV` and `R2`, with no servers or containers. What sets it apart from an ordinary tracker is that coordination of several agents lives in the same schema as the tickets. Issue leases and file claims hang off a ticket, and refused or forced claims are logged and ranked in a code heatmap. Agents on other machines and CI runners can take leases too, because the service is deployed and reachable over the network. People keep a normal web app with a backlog, kanban board, epics, sprints, a nested wiki with full-text search, flow metrics and a feedback widget. Connect Claude Code with `claude mcp add --transport http projektor "https://<host>/mcp/<workspace-id>"` and sign in with your web account; headless agents use a workspace token.
Who it is for. Teams where several agents edit one repository and need a shared tracker that keeps them apart by task and by file.
Good fit when
- Two agents edit neighbouring files and you need to see who holds which issue and which file
- A CI runner or an agent on another machine has to lease an issue in the same tracker
- The agent has to keep documentation in a nested wiki with full-text search
- A conflict heatmap should show which files agents collide in most often
Not a fit when
- You want a local tracker with no cloud: the app runs on Cloudflare in your own account
- You have no Cloudflare account and no way to put Cloudflare Access in front of the Worker
- A single agent in one repository is fine with a plain task list
Example request
Create a projektor issue for the login form bug, lease it and claim the files you will editMCP
- Transport
- http
- Authentication
- OAuth
Limitations
- The "Deploy to Cloudflare" button in the README deploys a release from before account sign-in existed, so such an instance needs a workspace token or a manual deploy for now.
- Put Cloudflare Access in front of the Worker before anyone logs in, otherwise the interface is open to anyone with the address.
- A Cloudflare account is required: the app uses `D1`, `KV` and `R2`, and the MCP tool count, 126 at the time of the README, changes between releases.
- Headless agents use a workspace token that must not be printed to logs or committed.
How to disable. Remove the MCP server with `claude mcp remove projektor`, and delete the Worker and its data in the Cloudflare dashboard together with the `D1` database.
FAQ
How do I deploy projektor?
Use the Deploy to Cloudflare button in the `projektor-deploy-example` repository, then put Cloudflare Access in front of the Worker. The deploy guide has the details.
How do I connect projektor to Claude Code?
Run `claude mcp add --transport http projektor "https://<host>/mcp/<workspace-id>"` and sign in with your account. Headless agents need a workspace token.
What are issue leases and file claims?
An agent leases an issue and claims the files it works on. A refused or forced claim goes into the log and the code heatmap.
How is projektor different from an ordinary tracker?
Agent coordination is stored in the same schema as tickets, so "which issue is this claim for" is a join. The service is deployed, so agents on other machines can take leases.
Is there a demo without deploying?
Yes, the README links a live demo on `workers.dev` and the docs sit on the project site.
Discussion
No questions or reviews yet.
Related
MCP serverstracker-mcp
traker-mcp
A Yandex Tracker MCP server with 187 tools and a skill: issues, queues, boards, sprints and automation over API v3, with a read-only mode
Jira MCP in Go: an agent searches issues by JQL, runs sprints, comments and worklogs and sees an issue's branches and pull requests on Cloud and Server
A free task manager with issues, sprints and docs and a built-in MCP server with about sixty tools.
Serverless notes on Cloudflare Workers with an MCP endpoint and a skill so agents can search and write your notes
Paca is a self-hosted task tracker where AI agents work as Scrum team members alongside people
A self-hosted task manager for human and agent collaboration through MCP
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| projektor (this card) | no | OAuth | 66 | |
| Seeder | no | API key | 126 | Seeder also manages tasks over MCP; projektor adds issue leases, file claims and a conflict heatmap for several agents. |
| Scrumboy | yes | API key | 450 | Scrumboy is built on sprints and metrics, while projektor coordinates parallel agents inside one issue schema. |
README in short
The README presents projektor as an AI-native tracker and wiki on Cloudflare, lists what makes its coordination different and leads to deploying, connecting Claude and reference docs on tools and workflows.