Kanzo Board
Give people and coding agents one lightweight board to pick up, move, and discuss tasks.
- Source repo
- emanuele-em/kanzo
- Stars
- ★ 1k
- Last updated
- today
- License
- AGPL-3.0
- Primary language
- Rust
- FA score
- 49/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex · Claude Code
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- An individual developer wants Claude Code or Codex to take work from a shared task list and report progress.
- Not a fit if
- Organizations that require team accounts and fine-grained access controls
- Teams that need a publicly hosted internet-facing board
- Source review
- 49/100 · Major gaps
What does this agent do, and when should you use it?
Kanzo combines a browser-based kanban board and an MCP server in a single program of about 4.5 MB, with task data stored in a local SQLite database under ~/kanzo. People manage projects and cards in the browser or through the kanzo CLI, while coding agents use MCP to take tasks, move cards, and comment. Agents identify a project from the Git remote of the repository they are working in. Boards can be shared live on a trusted network or synced between computers used at different times through a Git repository. It suits individuals and small teams who want human planning and agent work in one task flow; the documentation does not describe accounts or logins.
Running kanzo starts the local web board and opens http://localhost:4242; the first run creates ~/kanzo. People create, search, filter, and edit tasks in the browser, adding descriptions, comments, and image attachments. Agents connect through kanzo mcp or /mcp, match a project using the current repository's Git remote, take tasks, move cards to doing, and post progress comments. Optional kanzo share on and kanzo join <ip> let multiple computers use one board live; kanzo sync commits and pulls or pushes board data through a Git repository supplied by the user. kanzo export produces Markdown or JSON.
- An individual developer wants Claude Code or Codex to take work from a shared task list and report progress.
- A small team wants several computers and agents to update one kanban board over a trusted home or office network.
- A developer who alternates between a laptop and desktop wants to sync a local board through their own private Git repository.
- A Raspberry Pi owner wants to keep a board available on the local network and start it after reboot with
kanzo service install. - A user wants task data for documents or scripts and exports it as Markdown or JSON with
kanzo export.
How do you install or deploy this agent?
On macOS and Linux, including Raspberry Pi, use the installer script. Windows has a PowerShell installer. Cargo is another documented installation route. The README does not require an API key or paid service.
How do you use this agent?
Start the board and add Kanzo as an MCP server for your coding agent:
What are this agent's strengths and limitations?
- One program of about 4.5 MB provides both the web board and MCP server; the README reports about 4.5 MB idle memory use.
- Supports live sharing with
kanzo share onandkanzo join <ip>, plus Git-based sync using a repository you supply. - Agents can take, move, and comment on tasks through MCP; the README reports 10 to 60 tokens of plain text per agent reply.
- Stores tasks locally in SQLite and exports them as Markdown or JSON.
- Live sharing has no login mechanism; the README advises sharing only on a trusted network.
- Cross-machine sync requires a Git repository the user can push to; when both sides changed, the user must choose which version to keep.
- Agent integration requires an MCP client configuration; the documented examples are Claude Code and Codex.
- The README does not describe user accounts, team permissions, or fine-grained access controls.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Kanzo Board This agent | 49 · Major gaps | CLIFree | ★ 1k | today | Rust | Codex · Claude Code |
| AI Team OS | 81 · Good | Agent plugin / skillFree + model costs | ★ 374 | 9d ago | Python | Claude Code |
| Athena | 68 · Some gaps | IDE extensionFree + model costs | ★ 620 | today | Python | Claude Code |
| COMPASS Skills | 85 · Good | CLIFree | ★ 762 | 6d ago | Python | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
README states default bind 127.0.0.1, sharing is opt-in, no logins and warns to share only on trusted networks, giving thin least-privilege and external-effect coverage (1 each); but shared mode has no auth and MCP can write cards with no confirmation evidence (user_confirmation 1). Data flow only describes local ~/kanzo and optional git sync, not MCP transport or exposure details (data_flow_transparency 1). No sensitive-data handling notes (1). Dependencies are pinned and CI uses --locked, but no audit/vuln scanning evidence (dependency_security 1). Rollback limited to sync conflict choice and update, no version rollback (rollback 1). Attribution exists via repo/crates metadata but publisher unverified (source_attribution 1).
Cargo.toml pins versions and CI runs fmt/clippy/test on three OSes, good self-consistency (self_consistency 2); dependencies are common crates.io libs (dependency_availability 2); no failure-message/error-handling documentation (failure_messages 1).
README covers humans, agents, multi-machine sharing, Pi, VPN scenarios (audience_and_scenarios 2); capability boundaries only vaguely describe MCP tools without listing them (capability_boundaries 1); triggers rely on natural-language prompts with no precise spec (trigger_precision 1); cross-platform install and --home/port config are solid (environment_fit 2).
README is well-structured with command table and settings (information_architecture 2); multiple install paths with verification (install_notes 2); no versioned API/naming stability statement (naming_stability 1); rich examples and tips (examples_and_faq 2); known limitations limited to one line on no-login/trusted network (known_limitations 1); full AGPL-3.0 LICENSE matching Cargo (license 3); no CHANGELOG (versioning_changelog 1); maintenance responsibility unclear (maintenance_responsibility 1).
Outputs as board/Markdown/JSON export are usable (output_usability 2); adds marginal value for agent collaboration (marginal_value 2); size/resource claims have concrete numbers but are unverified, cost-benefit reasonable (cost_benefit 2).
Performance numbers cite a measurement environment but no benchmark scripts (claim_traceability 1); README partially corroborated by Cargo/CI (cross_source_corroboration 1); facts and inferences mostly separated but not explicitly labeled (fact_inference_separation 1).
- Shared mode has no authentication; README itself says there are no logins and to share only on trusted networks—add VPN/firewall before exposing publicly.
- The MCP server can write cards and comments; no evidence of confirmation or permission tiers for agent actions.
- Performance/size figures (4.5MB, 2ms, 10-60 tokens) are self-reported and not independently verified.
- Publisher identity is unverified and install uses curl|sh; verify GitHub attestation before running.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
What's new Share one board between computers. Run kanzo share on on one computer and kanzo join <address> on the others, or click the share icon next to New task. Everyone sees the same tasks and every agent works on them. Nothing to set up and nothing restarts. It works on your home network, or from anywhere over a VPN. Kanzo runs natively on Windows, with a PowerShell installer. Thanks @marcuslannister! Agents can hand a task to you. They mark it Human over MCP and leave it alone. The person icon next to New task, or h, shows only those. Thanks @eniewold! …View on GitHub ↗ -
Release notes excerpt
Install kanzo 0.2.1 Install prebuilt binaries via shell script ```sh curl --proto '=https' --tlsv1.2 -LsSf https://github.com/emanuele-em/kanzo/releases/download/v0.2.1/kanzo-installer.sh | sh ``` Download kanzo 0.2.1 | File | Platform | Checksum | |--------|----------|----------| | [kanzo-aarch64-apple-darwin.tar.xz](https://github.com/emanuele-em/kanzo/releases/download/v0.2.1/kanzo-aarch64-apple-darwin.tar.xz) | Apple Silicon macOS | [checksum](https://github.com/emanuele-em/kanzo/releases/download/v0.2.1/kanzo-aarch64-apple-darwin.tar.xz.sha256) | …View on GitHub ↗ -
Release notes excerpt
Install kanzo 0.2.0 Install prebuilt binaries via shell script ```sh curl --proto '=https' --tlsv1.2 -LsSf https://github.com/emanuele-em/kanzo/releases/download/v0.2.0/kanzo-installer.sh | sh ``` Download kanzo 0.2.0 | File | Platform | Checksum | |--------|----------|----------| | [kanzo-aarch64-apple-darwin.tar.xz](https://github.com/emanuele-em/kanzo/releases/download/v0.2.0/kanzo-aarch64-apple-darwin.tar.xz) | Apple Silicon macOS | [checksum](https://github.com/emanuele-em/kanzo/releases/download/v0.2.0/kanzo-aarch64-apple-darwin.tar.xz.sha256) | …View on GitHub ↗
FAQ
Does Kanzo require an API key or a paid model?
How does an agent find the right project?
Can I use the board on multiple computers?
Where is my board data stored?
~/kanzo, with the database in kanzo.db. Use --home <dir> or KANZO_HOME to choose another directory.