Retry failed webhooks
WORKING
4m 12s
Run multiple Claude Code agents in parallel, from one board.
Kanzai gives every card its own Claude Code agent, worktree and branch. Drag three cards into Development and three agents start side by side. The board shows which one is working, which one is waiting for you, and which one is ready for review.
Start using Kanzai for free. No account required.
macOS · Windows x86_64 · Linux AppImage x86_64 · Linux deb x86_64 · arm64 builds
A screenshot of the Kanzai app. Two projects sit in the tab bar. The Petshop board runs Backlog through Done, three cards are in Development, the pane on the right shows the selected session's transcript, and the project assistant is asking permission to move three cards to Ready.
THE PROBLEM
Several agents in several terminals are hard to track, and harder to review.
Running Claude Code in parallel is easy to start. Open three terminals, start three agents, and the work goes three times as fast. Until one of them stops to ask a question and you do not notice for twenty minutes.
Then you tab through windows to find out which agent is still working, which one is waiting, and which one is done. Two agents in the same checkout overwrite each other's files. Each finished diff has to be found, read and merged by hand, in the right order.
Kanzai puts each agent on a card. Every card runs in its own worktree on its own branch, so parallel agents cannot collide. The board shows at a glance which card needs you.
HOW KANZAI SOLVES IT
One card, one agent
Each card runs its own Claude Code agent, tied to the task it was given. The free version runs three agents at once; Unlimited lifts the cap.
A worktree per agent
Every card gets its own worktree and branch. Parallel agents change the same repo without touching each other's files.
See which one needs you
Each card shows whether its agent is working, waiting for an answer, or done. Finished work moves to Review, so nothing sits unnoticed in a background terminal.
Review each diff on its own
Open a card to read its diff, comment on individual lines, and send all your feedback back to that agent in one go.
Merge in the order you choose
Approved cards go into a merge queue. You set the order. If something fails, Kanzai stops and shows exactly what made it in.
The real Claude Code
Kanzai runs the Claude Code CLI you already use, with your Claude subscription. No API key. Codex, Grok Build and Antigravity run the same way; OpenCode may need an API key for its provider. Read the transcript, or switch to the terminal.
HOW TO RUN CLAUDE CODE IN PARALLEL
01
Install Kanzai
Download it for macOS, Windows or Linux. It is free and needs no account. Have Claude Code installed and signed in.
02
Write a card per task
Add your local repo, then write one card for each thing you want done. Keep each card to one change.
03
Drag them into Development
Every card you drag in starts its own Claude Code agent in its own worktree. Three cards, three agents, running side by side.
04
Answer, review, approve
Watch the board. Answer the card that is waiting, read the diffs that reach Review, and drag the good ones into Approved to merge.
- Native app for macOS, Windows and Linux
- Runs locally, with no account and your data in one file
- Runs the real Claude Code, Codex, Grok Build, Antigravity and OpenCode CLIs
- Use your Claude or ChatGPT plan, xAI login or Google sign-in; only OpenCode may need an API key
- Works with any Git host
Stop tabbing between terminals.
Run multiple Claude Code agents in parallel from one local board. Each gets its own worktree, and the board tells you which one needs you. Keep using the subscription you already pay for.
Download Kanzai free
Free to use · No account · No API key · No telemetry
The button picks x86_64 for Windows and Linux. On arm64 hardware, take the arm64 build.