Kata
Wes McKinney's local-first issue tracker built for AI-assisted software work. An agent-friendly CLI (JSON output, stable commands, predictable failure modes) paired with a human-facing TUI (browse, triage, edit, supervise). Both talk to the same local daemon and SQLite database.
Key Themes
The architectural bet: the issue ledger should be a local service adjacent to workspaces, not a database owned by each repository. A repo using kata gets only a small .kata.toml binding file; canonical state lives in KATA_HOME behind a daemon API. This keeps task state out of code history while giving agents a structured coordination layer.
Design priorities: - Agent ergonomics -- stable commands, JSON-first workflows, explicit workspace binding, search-before-create, idempotency keys, predictable exit codes - Human oversight -- TUI for browsing and supervising agent activity without reading raw JSON - Auditability -- append-only comments, event history, actor attribution, explicit destructive operations
Compared to Beads (Dolt-powered distributed graph tracker), kata is deliberately smaller: one daemon, one local store, one HTTP API, one TUI, and a narrow issue model. Where Beads offers distributed database semantics and federation, kata offers simplicity and teachability.
This connects to the broader task-tracking-for-agents category in Agentic Coding. When the Target Keeps Moving provides the framework for what to track (discovery vs. delivery); kata provides the mechanism. And Radical Accountability is McKinney's broader thesis -- the same person building kata is arguing that AI removes the excuse for mediocre software.
Critical Analysis
Kata makes smart trade-offs. Keeping state out of the git repo avoids polluting code history. Using a local daemon instead of a hosted service keeps the tool fast and dependency-free. The narrow issue model (no project management, no workflow engine) is a feature -- it's easy to teach to agents.
The risk: "local-first" means no built-in collaboration. A shared server mode is planned but not yet built. For solo agent-assisted work, kata is ideal. For teams, the lack of multi-user coordination may push people to Beads or traditional issue trackers. The question is whether kata will ship the shared mode before the window closes.
Sources: kata Last updated: 2026-05-14