SkillFoxx

kata CLI

kata is a local issue tracker for coding agents: the agent creates, claims and closes issues with evidence, humans watch in a TUI

  • 518 ★ on the repository
  • Last commit
CLI

Extended access

Why this level

  • Runs a background daemon and kata cron schedules that execute jobs on triggers without asking
  • Keeps issue and agent comment content in a local database and sends it to a shared server when a remote daemon or federation is enabled
All reasons and checks

kenn-io/kataAuthor:kenn-io

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add cli/kata

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • Runs a background daemon and kata cron schedules that execute jobs on triggers without asking
  • Keeps issue and agent comment content in a local database and sends it to a shared server when a remote daemon or federation is enabled

This is third-party code. Review the repository files before installing.

What it does

kata is an issue tracker written in Go for coding agents and the humans steering them. Agents need a place where work survives context compaction and multiple workers, and a chat thread or a markdown to-do list is not that. kata gives a local task ledger: the agent creates, claims, relates and closes issues from the command line and attaches evidence of what was done when closing, for example --test with a check. State lives in a local SQLite database by default, so the repository stays clean and no hosted tracker is required. Humans follow the same issues in the kata tui terminal UI or in a browser through kata ui. kata cron runs recurring jobs and workflows on triggers with per-run evidence, and kata notify flags an issue for a person or agent with a reason. Helpers get an identity of their own: KATA_TEAMMATE attributes their comments and issues, and KATA_INBOX_USER opens their inbox. A reply to a comment can be backed by evidence: confirm, refute or supersede. For teams there is a remote daemon and federation with Postgres through KATA_DSN. kata mcp serve starts an MCP server over stdio or Streamable HTTP, and for Go there is an embeddable library.

Who it is for. Developers who give agents multi-step work and want to see what was done, by whom and with what evidence, without a hosted tracker.

Good fit when

  • The agent loses its to-do list when context is compacted and issues should live outside the chat
  • Several agents or subagents work on a repository and each needs its own identity and inbox
  • Closing an issue must come with a check: kata close --done --test records what confirms the work
  • You need a task MCP server for Claude Code, Cursor or Codex: kata mcp serve

Not a fit when

  • The team already tracks work in Jira or GitHub Issues and does not want a second tracker
  • You need only a to-do list for one session: the agent's built-in plan is enough
  • You want a ready hosted tracker with no daemon of your own: shared mode needs a remote daemon deployed

Example request

Create a kata issue for the login race, claim it, and after the fix close it with the test command as evidence

Limitations

  • Building from source needs Go 1.27.4 or newer. A go install build does not include the compiled browser bundle, so for kata ui install a release binary or build a clone with make install.
  • Project bridges, team visibility and instruction install for hookless harnesses need Kata 0.19.0 or newer.
  • Shared work needs a remote daemon or federation, and Postgres for a shared daemon is configured through KATA_DSN.
  • Project limits in the MCP server need a daemon with API version 0.35.0 or newer.

How to disable. A Homebrew install is removed with brew uninstall kata, and a release archive build is removed by hand. Remove the MCP server from the agent's configuration. The task database lives locally in the kata data directory and is deleted by hand. The agent instruction installer is removed with kata agent-hook instructions uninstall.

FAQ

How does kata work with coding agents?

The agent finds issues with kata search, claims an open one with kata claim or creates a new one with kata create, comments with findings and closes the work with kata close and evidence. The loop is in the README, and kata quickstart covers subagents, inboxes and schedules.

Do I need a server-side tracker?

No. By default state lives in a local SQLite database and the repository stays clean. For shared work you can enable a remote daemon or federation, and a shared daemon can use Postgres through KATA_DSN.

How do I connect kata as an MCP server?

Run kata mcp serve over stdio or Streamable HTTP. By default the server is bound to the workspace's project, a project allowlist is set with --projects, and --all opens every project visible to the daemon.

How do I install kata?

On macOS with brew install kata, on Linux with the script curl -fsSL https://katatracker.com/install.sh | bash, on Windows with the PowerShell script from the docs. The script verifies the release archive against SHA256SUMS. Check with kata version.

What is evidence when closing an issue?

A record of what was done and by whom: a message and a check, for example kata close <id> --done --message "..." --test <check>. If the work is not verified, label the issue needs-review with a comment.

Discussion

No questions or reviews yet.

CLI tools
Russian stack

Command-line client and MCP server for the Kaiten tracker: cards, comments, checklists and files from the terminal or an agent

CLIBroad access5

A self-hosted issue tracker and wiki on Cloudflare with MCP: 126 tools, issue leases and file claims for several agents

MCP serverBroad access66

A small issue tracker for AI agents: tasks, epics and dependencies live in a project SQLite file, with a local board.

CLIMinimal access88

A task tracker for agents: tasks with context and result live in JSONL beside the code and outlive sessions

CLIMinimal access410

CodeKanban gathers terminals and Claude Code and Codex sessions on one page, and codekanban-cli drives them from an agent

CLIBroad access227

A terminal kanban board in Python that Claude Code and other agents can read and change through a CLI or MCP

CLIExtended access289

README in short

The README explains the idea of a tracker for agents, lists capabilities from the CLI to federation and MCP and gives installation for macOS, Linux and Windows. A five-step agent work loop and a command quickstart follow.

README badge

Report a problem