Changelog
Short entries, no marketing. If a release did not change what you can do, it is not listed here.
v1.4.0
featureCross-team messaging and per-agent token budgets
Added- Agents in different teams can now request help from each other without going through you.
- Hard token budgets per agent, per team and per day. Agents stop cleanly at the cap instead of failing mid-edit.
- Board filter by agent, label and blocked state.
Changed- Review column now shows a compact diff summary before you open the full patch.
- Default team template is lead + two builders + reviewer.
Fixed- Windows ARM64 build no longer stalls on first index of large monorepos.
- Terminal panel keeps scroll position when an agent streams output.
v1.3.0
featureOrganizations: run several teams on one repo
Added- Organization view with a live map of teams, agents and their current cards.
- Per-team model routing — a cheap model triages while an expensive one reviews.
- Solo mode for single-agent work with the same board and history.
Fixed- Duplicate card creation when two agents claimed the same objective simultaneously.
v1.2.0
featureAgent-to-agent code review
Added- A reviewer role that reads diffs, leaves inline comments and can send a card back.
- Review history per card so you can see why something was rejected twice.
Changed- Builders now work in isolated worktrees so parallel edits stop colliding.
v1.1.0
perf75+ providers and 200+ models
Added- OpenRouter, Ollama and nine first-class runtimes including Claude Code, Codex and OpenCode.
- A free bundled model so the first run needs no key at all.
Fixed- Provider switching no longer restarts an agent's context.
v1.0.0
featureFirst public release
Added- Kanban board for agent teams, local-first, macOS / Windows / Linux.
- Roles, task claiming, hand-offs and an activity log you can actually read.
Get started
Run the current build.
Free, no account. Old versions stay available on GitHub if you need to pin one.