Your AI coworker finally gets its own Jira tickets

What it does Paca is a self-hosted project management platform that reimagines Scrum tools for mixed teams of humans and AI agents. Instead of bolting a chatbot onto a dashboard, it assigns AI agents to sprints, lets them pick up tasks from a shared Scrumban board, and co-author BDD specs and system design documents alongside human teammates. The core is lightweight and everything else — workflows, fields, board layouts, even agent behavior — is driven by project-level configuration files or WASM plugins.

The interesting bit The unusual angle is treating AI agents as first-class team members rather than peripheral automation. Agents run in isolated sandbox containers via the OpenHands SDK, join sprint planning, update task status in real time, and probe complexity alongside humans. Paca also exposes an MCP server and a Claude Code /paca skill, so external AI tools can read and write the workspace directly without scraping or custom API glue.

Key highlights

• AI agents appear on the same real-time Scrumban board as humans, pulling tasks and posting updates via Socket.IO
• Extensible via WASM plugins with capability-based permissions; backend plugins can be written in Go, Rust, or anything compiling to WASM
• Ships with an MCP server (@paca-ai/paca-mcp) and a Claude Code slash command for managing tasks without leaving the editor
• In-app AI chat for plain-English project planning, plus activity diffs with one-click revert
• Self-hosted, Apache 2.0, with a minimal core that scales down by disabling services like the AI agent or MinIO

Verdict Worth a look if you want AI agents inside your Scrum workflow rather than bolted onto the side of it. Skip it if you need a mature, battle-tested enterprise tracker today — this is a v0.4.0 project with just over a thousand stars and a bold thesis to prove.