Paca: The AI-Native Open-Source Alternative to Jira That Treats Agents as Teammates

By Prahlad Menon Published 2026-06-14 2 min read

Most “AI-powered” project management tools bolt a chatbot onto the side. Ask it questions, get summaries, maybe auto-generate a few fields. The AI is a helper, not a teammate.

Paca takes a fundamentally different approach: AI agents are first-class Scrum team members. They’re assigned to sprints. They appear on the board. They pick up tasks, update statuses, write BDD specs, and collaborate on system design documents — just like any human teammate.

Why This Matters

The Jira/ClickUp/Monday model treats AI as automation infrastructure — pipelines that generate output in isolation. Paca is built on a different insight: complex work requires teams, not pipelines. AI agents should participate in the Scrum process itself.

In Paca:

Agents are assigned to sprints alongside humans
They pull tasks from the backlog and update status in real-time
They co-author Gherkin scenarios with Product Owners and BAs
They contribute to System Design Documents, keeping context visible to everyone
They probe, sense, and respond to emerging complexity

This isn’t automation bolted on. It’s genuine collaboration.

The Stack

Layer	Tech
Frontend	React + TanStack Start + shadcn/ui
API	Go + Gin
Realtime	Node.js + Socket.IO
AI Agents	Python + FastAPI + OpenHands SDK
Storage	PostgreSQL + Valkey
Plugins	WebAssembly (WASM) sandbox

Everything runs via a single docker compose up -d. Self-hosted, Apache 2.0 licensed, your data never leaves your infrastructure.

Key Features

MCP Server for Any AI Agent

Paca ships an MCP (Model Context Protocol) server that gives Claude, custom agents, or any MCP-compatible tool structured access to your workspace. No scraping, no custom integration work.

{
  "mcpServers": {
    "paca": {
      "command": "npx",
      "args": ["-y", "@paca-ai/paca-mcp"],
      "env": {
        "PACA_API_KEY": "your-api-key",
        "PACA_API_URL": "http://localhost:8080"
      }
    }
  }
}

Now Claude can create tasks, manage sprints, update documents — all through natural language.

Claude Code Integration

If you use Claude Code, install the Paca skill set and manage your workspace through slash commands:

curl -fsSL https://raw.githubusercontent.com/Paca-AI/paca/master/scripts/install-claude-skill.sh | bash

Then /paca create a task for implementing OAuth in sprint 3 just works.

Command	What it does
`/paca <request>`	General task/doc/sprint operations
`/paca-epic <requirements>`	Turn requirements into epic + stories + spec
`/paca-breakdown <task>`	Decompose into sub-tasks
`/paca-sprint`	Plan sprint from backlog
`/paca-do <task>`	Execute task and update status

WASM Plugin System

Extend or replace any part of Paca via plugins. Backend plugins compile to WebAssembly (write in Go, Rust, AssemblyScript), frontend plugins are standard module bundles. Plugins run in a sandboxed environment with capability-based permissions.

Browse and install community plugins directly from the Plugin Marketplace in the UI — no CLI required.

OpenHands-Powered Agents

AI agents run on the OpenHands SDK. Each agent executes inside its own isolated sandbox container, so your host environment is never touched.

The Plan-Act-Check-Adapt Loop

Paca structures collaboration around four phases that mirror Scrum and the scientific method:

Plan → Act → Check → Adapt
  ↑                      |
  └──────────────────────┘

Plan: POs, BAs, and AI agents collaboratively refine the backlog
Act: Sprint is live — humans and agents pull tasks and execute
Check: QA agents run automated verification, humans review
Adapt: Data informs the next cycle, team retrospects together

How It Compares

Feature	Jira/ClickUp/Monday	Paca
AI integration	Chatbot add-ons	First-class Scrum teammates
Collaboration	Human-only default	Human + AI on same board
Hosting	Vendor cloud	Self-hosted
Cost	$8-20+/seat/month	Free forever
Customization	Enterprise tiers	Config + plugins
Source	Proprietary	Apache 2.0

Getting Started

One command:

curl -fsSL https://github.com/Paca-AI/paca/releases/latest/download/install.sh | bash

Or manual setup:

mkdir paca && cd paca
curl -fsSL https://github.com/Paca-AI/paca/releases/latest/download/docker-compose.yml -o docker-compose.yml

# Create .env with your secrets
docker compose up -d

Open http://localhost, log in as admin.

The Name

“Paca” is a pun on the Japanese word “Baka” (ばか) — “silly.” The team jokingly called their AI assistants “silly” when they hallucinated. Building a free open-source alternative to multi-billion-dollar tools might also seem silly.

But here we are: 730 stars, Apache 2.0, and a genuinely different take on how AI should fit into software development teams.

Links:

GitHub: Paca-AI/paca
License: Apache 2.0
Stars: 730+ ⭐

If you’re tired of Jira’s bloat and want AI that actually participates in your process rather than just answering questions about it, Paca is worth a look.