Player2 Kanban: Local-first Project Management for Agentic Engineering
A lightweight Kanban board that keeps AI agents aligned with real engineering discipline
The Problem
AI coding agents are powerful, but they're also unconstrained. Without guardrails, vague tickets are created, acceptance criteria is skipped, and engineering discipline is ignored. Popular project management tools like Jira and Linear live in the cloud, and often cover massive groups of teams. This means context switching, learning curves, unknown field customization, and data leaving the session.
You need a tool that keeps agents accountable to engineering standards while living locally, on your machine, under your control.
Origins
Player2 Kanban is built on Thomas Caraccia's taskboard, an elegant single-binary Kanban board with an embedded web UI and no cloud dependency. I discovered taskboard while looking for a lightweight alternative to Jira and Linear, used it, and quickly ran into the walls of my own requirements.
Rather than building from scratch, I forked taskboard and rebuilt the internals to support multiple projects, GitHub synchronization, AI agent integration via MCP, and strict mode enforcement. The core architecture supports a local-first, single binary, embedded web UI.
Credit to Thomas for the foundation and inspiration. Player2 Kanban stands on his open-source development.
Core Features
Bidirectional GitHub Sync
Projects link to GitHub repositories. Tickets sync to issues and back automatically. Metadata is stored in hidden HTML comments so your issue body stays clean. Background workers handle sync with exponential backoff and rate limit awareness.
AI Agent Integration via MCP
AI agents (Cursor, Claude, Copilot, Windsurf, Gemini) connect via player2-kanban mcp and get access to 20+ tools for managing projects, tickets, and boards. Tool descriptions embed hierarchy guidance so agents don't create vague tickets.
Strict Mode Enforcement
When enabled, non-draft tickets require a User Story ('As a... I want... So that...') and Acceptance Criteria (Given/When/Then). This is enforced the moment an agent attempts to push data to the system, and cannot be bypassed. This is the most impactful feature for maintaining a quality, agentic workflow.
CLI for Everything
Every operation available in the web UI is also available from the command line. Create tickets with strict mode fields, check sync status, manage teams and projects, install hooks, all without opening a browser.
Agent Configuration
One command drops rule files into your repo that teach AI agents your project's lifecycle protocol: check for tickets before coding, move tickets through statuses, sync on completion. Works with Cursor, Claude, Gemini, Windsurf, Copilot, and Codex.
Git Hooks for Automation
Pre-push and post-merge hooks auto-sync the board so it stays current without manual intervention. Keep your project management and version control in lockstep.
Who It's For
Developers and small teams who use AI coding agents (Cursor, Claude Code, Copilot, Windsurf) for real implementation work and need those agents to follow a structured workflow
Teams working across multiple repositories who want project management that lives locally rather than in a SaaS tool
Organizations that care about strong product and engineering discipline and observability require valid work items with proper acceptance criteria, and meaningful context before work happens.
Technical Foundation
Key Design Decisions
Single Binary, Embedded Assets
The React frontend compiles into the Go binary via embed.FS. No Docker, no separate frontend server, no runtime dependencies. npm install -g player2-kanban gives you a working tool in seconds.
Local-First with SQLite
No account creation, no network dependency, no data leaving the machine. The database lives in your OS config directory. This is a developer tool, built to enable people and teams to enforce structure in their effort to build great products.
MCP Over REST
REST is available (and the web UI uses it), but MCP is the native interface for AI agents. It provides typed tool definitions with descriptions that guide agent behavior, which a generic REST API can't do.
Strict Mode is a gate, instead of a mere suggestion.
There are many tools that make standards optional. In an agentic workflow, optional work gets ignored. Strict mode makes user story and acceptance criteria a hard requirement at the system's front door, and agents that skip them are not enabled to bypass these guardrails.
Tech Stack
Current Status
The project is open source on GitHub at github.com/Tristan578/taskboard and published on npm as player2-kanban.
CI enforces 75%+ test coverage, security scanning via gosec, and linting. Binaries are built for Linux, macOS, and Windows (amd64 and arm64) via GitHub Actions on every tagged release.
Get Started
Player2 Kanban is ready to install. Install globally, initialize a project, connect your GitHub repos, and let your agents work within your engineering standards.
Common questions
What is Player2 Kanban?
Player2 Kanban is Tristan J. Nolan's open-source, local-first Kanban board: a single Go binary with GitHub sync and an MCP server, so AI coding agents can read and update the board while a strict mode enforces user stories and acceptance criteria at the API level.