Paca: Scrum task tracker with MCP for agents
Paca
Paca is a self-hosted task tracker where AI agents work as Scrum team members alongside people
- 1,997 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- The API key lets the agent write to the tracker, including deleting projects and tasks
- A role with broad rights opens every project on the instance to the agent
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/pacaDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The API key lets the agent write to the tracker, including deleting projects and tasks
- A role with broad rights opens every project on the instance to the agent
This is third-party code. Review the repository files before installing.
What it does
Paca is a free open-source task tracker you host yourself, positioned as a replacement for Jira, Trello, ClickUp and Monday. AI agents work in it as Scrum team members: they join sprints, pick tasks from the board, write BDD scenarios and update statuses. The @paca-ai/paca-mcp package runs through npx and connects any MCP client to the tracker. The agent gets tools for projects, tasks, sprints, documents, members and roles, task types and statuses, views and custom fields, attachments, comments and activity history. Sample requests: list the active sprints of a project, create a task for OAuth and attach it to sprint 3, add a comment with a progress report to a task. Plugins installed in Paca register their own tools while the server runs. The repository also installs a set of `/paca` commands for Claude Code, Gemini CLI and Cursor: turn requirements into an epic with stories, estimate tasks in story points, plan a sprint against team capacity, execute a task and update the docs. The commands are fetched from your own Paca instance, so their version matches the server.
Who it is for. For teams that run sprints in their own tracker and want Claude Code, Cursor or Gemini CLI agents to work on the board and documents directly.
Good fit when
- Break requirements into an epic with child stories and put them on the board
- Plan a sprint: the agent reads the backlog, estimates and goals and proposes the scope
- Record progress on a task: change status, leave a comment, attach a result
- Keep specs and design documents in Paca Docs without leaving the editor
- Connect several agents with different roles and limited rights to one board
Not a fit when
- Your team lives in Jira or Linear and is not moving: those systems have their own MCP servers
- You have no server with Docker to deploy Paca on
- You only need a personal to-do list without sprints and roles: a multi-service platform with PostgreSQL and Valkey is heavy for that
Example request
Show the active sprints of the Platform project, create a task for OAuth sign-in and add it to sprint 3.MCP
- Transport
- stdio
- Authentication
- API key
Environment variables
PACA_API_KEYrequired, secret- API key from the Paca profile settings
PACA_API_URL- Paca API address, http://localhost:8080 by default
Limitations
The package itself stores nothing: it needs a running Paca instance and an API key from your profile settings. The Paca server and the package run on your infrastructure.
The platform consists of a Go API, a real-time events service, a web interface, a PostgreSQL database and a Valkey cache, so it needs a separate server with Docker.
Paca can also run in Kubernetes through Helm. The chart requires publicUrl and secrets to be set, as they have no defaults.
How to disable. Remove the server from Claude Code with `claude mcp remove paca`, and in Claude Desktop delete the paca entry from the claude_desktop_config.json file. Delete an API key you no longer need in the Paca profile.
FAQ
How do I connect Paca to Claude Code?
Run `claude mcp add paca --env PACA_API_KEY=<key> --env PACA_API_URL=<url> -- npx -y @paca-ai/paca-mcp`. The key is created in the Paca profile settings under API Keys. For Claude Desktop add a paca entry to the claude_desktop_config.json file and restart the app.
How is Paca different from Jira and Trello?
Paca runs on your own server, is licensed under Apache 2.0 and charges no per-seat fee. Agents are assigned to sprints and pick tasks from the board like people do. Workflow, fields and board layout are set through configuration, and extensions are written as WebAssembly plugins.
What tools does the Paca MCP server provide?
Examples include list_projects, create_task, complete_sprint and add_task_comment. The set also has tools that delete projects and tasks. The full list is in the repository file apps/mcp/ALL_TOOLS.md.
Can I limit what an agent may do in Paca?
Yes. Access in Paca comes from roles written as JSON policies that allow or deny actions on resources down to a single task, sprint or document. Without a role there is no access, and an explicit deny always beats an allow. Role edits apply from the next request.
How do I install the /paca commands for Claude Code?
Set your Paca address in the PACA_API_URL variable and run the repository script install-paca-skills.sh through curl and bash. It downloads the commands from your instance and places them in the Claude Code, Gemini CLI and Cursor directories. The guided connection setup runs with `/paca-setup`.
Discussion
No questions or reviews yet.
Related
MCP serversJira 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
Scrumboy is a self-hosted Go task tracker with kanban boards, sprints and metrics that includes an MCP server for Cursor and Claude Code
A spec-driven delivery skill: PRDs, epics, tasks, GitHub Issues sync and parallel agents in git worktrees
A self-hosted task manager for human and agent collaboration through MCP
The official Basecamp CLI, with plugins for Claude Code and Codex, lets you run projects and todos from the terminal or an agent
A task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Paca: Scrum task tracker with MCP for agents (this entry) | yes | API key | 1,997 | |
| Scrumboy | yes | API key | 448 | also a self-hosted tracker with sprints and MCP, written in Go and centered on kanban boards |
| Linear MCP Server | no | API key | 346 | connects the hosted Linear, while Paca is itself the tracker and runs on your server |
| Taskmaster | no | not required | 28,277 | manages development tasks of one repository for the agent, with no team board or roles |
README in short
The core idea: agents and people work as one Scrum team. The P-A-C-A cycle follows Scrum and the scientific method: plan, act, check, adapt. The core is small, and the rest comes from configuration and plugins from the in-app marketplace.