agent-kanban
Local-first kanban for AI-agent workflows. SQLite, FastAPI, no auth, no cloud.
Agenten stärken
Installation vom Herausgeber. Kein PlugMCP-Endpunkt.
Host a copy with our keys and logs.
Noch kein paketiertes Release in der Registry. Folgen Sie der README des Repositorys.
Remote-URL oder ein Befehl auf Ihrem Rechner? Worin sich die beiden Installationen unterscheiden
Auf PlugMCP ausführen
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
StartenVom Herausgeber
Local-first kanban for AI-agent workflows. Drag a task to Approved and your AI coding agent picks it up via MCP and drives it through analyst → in_progress → testing on its own. MCP for Claude Code & Cline, OpenAPI for opencode / Open WebUI / Ollama / any function-calling LLM. SQLite + FastAPI, runs on localhost.
A self-hosted kanban board designed to be driven by AI coding agents (Claude Code, Cline, opencode, Open WebUI, any function-calling LLM). Drag-drop in browser, REST + MCP for agents. Multi-project, light/dark theme, markdown plan-file import, automation rules.

Toggle theme with t, cycle profiles with p. Or pin a theme/profile via URL: ?theme=light&profile=cyberpunk.
/p/{slug}), each with optional Claude Code-project directory binding.t), density toggle (d), collapsible columns, sidebar./, n, r, \, Esc)..md into kanban_data/inbox/, get a card in 5 sec. Point KANBAN_INBOX_DIR at any folder you like — even ~/.claude/projects/<proj>/memory/inbox/ for Claude session captures.## Section → cards in backlog with section-name in description.kanban_data/rules.json): task_moved, task_idle, blockers_done, task_count_in_status → move, comment, assign, set priority, archive, run a command. Polling rules act once per stay in a column; WIP limits, a pause switch. Hot-reloaded.examples/agent-launcher/): card lands in Approved → headless Claude Code works it to Testing, with a timeout, one run per card, max_concurrent / max_runs, and the card blocked with a reason when a run goes wrong.kanban_data/webhooks.json): POST to Slack / Telegram / any HTTP endpoint on task_created/task_moved/task_commented/task_updated. 3 formats: generic, slack, telegram.kanban_mcp/) with 16 tools for Claude Code / Cline, strict arguments and compact answers; agents can't close cards themselves (configurable). python -m kanban_mcp.connect <dir> --project <id> writes the .mcp.json + CLAUDE.md block.X-Kanban-Actor./t/T-042.5 minutes from clone to an AI agent doing your work for you: see QUICKSTART.md.
Bare minimum (one command after install):
# Install uv once: `curl -LsSf https://astral.sh/uv/install.sh | sh` or `brew install uv`
git clone https://github.com/<your-user>/agent-kanban.git
cd agent-kanban
uv run python -m kanban_ui # http://localhost:7777
uv run auto-creates a .venv/, installs deps from pyproject.toml (~5 sec on a warm cache), and launches the server.
python3.12 -m venv .venv
.venv/bin/pip install -r requirements.txt
.venv/bin/python -m kanban_ui # http://localhost:7777
requirements.txt is kept in sync with pyproject.toml for Dependabot and pip-only environments.
Optional — auto-start on login (macOS):
bash scripts/install_launchd.sh install # macOS launchd
bash scripts/install_launchd.sh status
bash scripts/install_launchd.sh uninstall
Logs: ~/Library/Logs/agent-kanban/{stdout,stderr}.log.
Pick the integration matching your agent:
The kanban serves MCP over stdio (python -m kanban_mcp, for Claude Code / Cline), streamable HTTP at /mcp (Cursor, new Cline, MCP Inspector) and SSE at /sse for clients that only speak the older transport.
Static OpenAPI schema: docs/openapi.yaml. Interactive Swagger: http://localhost:7777/docs.
Real-world flows: docs/USECASES.md — 11 use cases (solo dev, team Slack, legacy import, multi-project, agent session, auto-launch, etc.).
agent-kanban/
├── kanban_store/ SQLite store + schema (5 migrations) + maintenance CLI
├── kanban_ui/ FastAPI web UI + automation/{events,inbox,rules,webhooks,plan_md}
│ └── static/ index.html · styles.css · app.js · vendor/Sortable
├── kanban_mcp/ MCP server (16 tools) + connect helper
├── seed/ generic example tasks
├── examples/ llm-tool-calling demo scripts
├── docs/ INTEGRATION.md, openapi.yaml, screenshots
├── tests/ pytest tests
├── scripts/ install_launchd.sh
├── snapshots/ JSON snapshots (gitignored except .gitkeep)
└── kanban_data/ runtime DB, inbox/, rules.json (gitignored)
python -m kanban_store.maintenance doctor # health report, read-only
python -m kanban_store.maintenance dedupe-history # dry run; --apply to delete (backup first)
python -m kanban_store.maintenance restore-archived # done cards an old rule relabelled "cancelled"
python -m kanban_store.maintenance vacuum # reclaim space (stop the server first)
Upgrading from 0.1.x: see CHANGELOG.md.
Local-first kanban for AI-agent workflows. Drag a task to Approved and your AI coding agent picks it up via MCP and drives it through analyst → in_progress → testing on its own. MCP for Claude Code &
Übersicht
Local-first kanban for AI-agent workflows. Drag a task to Approved and your AI coding agent picks it up via MCP and drives it through analyst → in_progress → testing on its own. MCP for Claude Code & Cline, OpenAPI for opencode / Open WebUI / Ollama / any function-calling LLM. SQLite + FastAPI, runs on localhost.
Tools
Noch keine Tool-Liste. Remote-Server erhalten eine nach dem ersten Health-Check.
FAQ
Kopieren Sie auf dieser Seite die stdio-Konfiguration für Cursor, Claude Desktop, Claude Code oder VS Code. Sie nutzt die Installation des Herausgebers, keinen PlugMCP-Proxy.
In diesem Eintrag ist keine erforderliche Umgebungsvariable aufgeführt.
Agent Kanban läuft lokal über stdio.
Zuletzt geprüft: 2026-10-28. Local-first kanban for AI-agent workflows. Drag a task to Approved and your AI coding agent picks it up via MCP and drives it through analyst → in_progress → testing on its own. MCP for Claude Code & Cline, OpenAPI for opencode / Open WebUI / Ollama / any function-calling LLM. SQLite + FastAPI, runs on localhost.