Task management built for AI agents — not against them
The open-source kanban board that speaks your agent's language. CLI-first, API-native, built for the age of agentic development.
Your AI agent can write code in seconds.
But managing tasks? That's still 6 API calls and a prayer.
Every project management tool was built for humans clicking buttons. Your agent needs an API-native system that treats task management as code — not an afterthought.
6+ curl commands to move a task through its lifecycle. Status update, timer start, agent status, timer stop, completion comment, status done. Every. Single. Time.
Jira, Linear, Notion — beautiful UIs for humans. Hostile APIs for agents. Your AI shouldn't need a browser to update a task.
Cloud-only, proprietary formats, pay-per-seat. Your task data locked behind someone else's API. No local dev, no file access, no control.
vk begin and vk done replace 6 API calls with 2 commands.
# Start working
curl -X PATCH .../tasks/<id> \
-d '{"status":"in-progress"}'
curl -X POST .../tasks/<id>/time/start
curl -X POST .../agent/status \
-d '{"status":"working"}'
# Finish up
curl -X POST .../tasks/<id>/time/stop
curl -X PATCH .../tasks/<id> \
-d '{"status":"done"}'
curl -X POST .../tasks/<id>/comments \
-d '{"text":"summary"}'
❯ vk begin US-42
✓ Status → in-progress
✓ Timer started
✓ Agent status → working
❯ vk done US-42 "Added OAuth2"
✓ Timer stopped (47m 23s)
✓ Status → done
✓ Comment added
✓ Agent status → idle
"Automate everything you do twice."— Boris Cherny, creator of Claude Code
18 CLI subcommands, full REST API, MCP integration, and a beautiful UI for when you want to look at the board yourself.
Built for AI agents from day one — not retrofitted onto a human tool.
vk begin, vk done, and 18 subcommands for full lifecycle control.
Automatic timer management. Start, stop, manual entries — all from CLI or API.
Plan before you execute. Subtasks, dependencies, and verification checklists.
Done criteria that enforce quality. Every task has a definition of done.
Bidirectional, label-based. Import issues, push status changes back.
Real-time, filterable chronological feed. Know what every agent is doing.
Model Context Protocol for Claude Desktop and any MCP-compatible assistant.
Full lifecycle management. If it can make HTTP calls, it can drive the board.
Auto-generated standup reports. Completed, in-progress, blocked, upcoming.
Time-boxed iterations with auto-archive, velocity tracking, and sprint filtering.
API key auth, JWT, CORS, rate limiting, CSP headers, XSS prevention. Production-ready.
Veritas Kanban is the only task board built from the ground up for AI agent workflows.
| Feature | Veritas Kanban | Jira | Linear | Plane | Planka |
|---|---|---|---|---|---|
| Open source | ✅ MIT | ❌ | ❌ | ⚠️ AGPL | ⚠️ Fair Use |
| Local-first | ✅ | ❌ | ❌ | ⚠️ Self-host | ⚠️ Self-host |
| AI agent orchestration | ✅ Native | ❌ | ❌ | ❌ | ❌ |
| MCP server | ✅ | ❌ | ❌ | ❌ | ❌ |
| CLI | ✅ | ❌ | ✅ | ❌ | ❌ |
| Git worktree integration | ✅ | ❌ | ❌ | ❌ | ❌ |
| Code review built-in | ✅ | ❌ | ❌ | ❌ | ❌ |
| Markdown file storage | ✅ | ❌ | ❌ | ❌ | ❌ |
| No database required | ✅ | ❌ | ❌ | ❌ | ❌ |
| Time tracking | ✅ | ⚠️ Addon | ❌ | ✅ | ❌ |
| REST API | ✅ | ✅ | ✅ | ✅ | ✅ |
| Free forever | ✅ | ❌ | ❌ | ⚠️ Limits | ✅ |
vk begin and vk done replace 6 API calls with 2 commands. Plus vk block and vk unblock.
Full time management from the terminal. Start, stop, manual entries, and summaries with vk time.
Add comments and manage agent presence from the terminal. Full lifecycle without touching a browser.
5 new command modules covering workflows, time tracking, comments, agent status, and project management.
No database required. File-based storage means your tasks are plain markdown you can grep, diff, and version control.
React 19 + TypeScript
Vite 7 + Tailwind CSS
shadcn/ui components
Node.js + Express
WebSocket real-time
TypeScript strict mode
Markdown + YAML frontmatter
No database required
Human-readable files
1,143+ unit tests (Vitest)
19 E2E tests (Playwright)
CI/CD with GitHub Actions
Clone, install, run. The board auto-seeds with example tasks so you can explore immediately.
Or ask your AI agent: "Clone and set up veritas-kanban locally."