Run every coding agent from one desktop
The open-source autonomous coding workspace for local and cloud agents. Give it an issue—Zuse plans, codes, tests, and prepares the pull request with the subscriptions you already pay for.
Free in beta · no token markup
What should we build in forkzero?
Bring the subscriptions you already pay for. Zuse orchestrates them — no token markup.
Review
From changed files to a verified pull request
Inspect the complete branch, commit deliberately, push when it is ready, and send failed checks back to the active agent.
Review every changed file
See the branch diff, changed files, and line totals before anything leaves your machine.
Branch review
Branch
agent/checkout-recovery
Changed files
9 modified · 2 added · 1 deleted
Branch diff
+182 additions · −47 deletions
Uncommitted work
3 files ready to review
Review range ready
Compared with origin/main
Inline diff review before push
Review additions, deletions, files, and agent findings without leaving Zuse.
Codex review: Retry only after the persisted checkout state is restored.
Commit selected files & push
Choose exactly which files enter the commit, write the message, and push the branch when it is ready.
Send failing checks to the agent
Collect failed GitHub Actions logs and attach them to the agent thread so it can diagnose and fix the branch.
checkout-e2e
GitHub Actions · failed
FAIL tests/auth-recovery.spec.ts
Expected redirect to /dashboard
Received /login?expired=true
36 log lines collected
Agent thread
agent/fix-auth
Failed check attached
failing-checks-0821.txt · 36 lines
Diagnosing session redirect…
Browser verification, built in
Let an agent drive a real browser and verify the flow inside Zuse
Import browser sessions
Copy valid cookies from a local browser profile into Zuse's built-in browser. Passwords are never imported.
Local profiles
Zuse browser
Session ready
3 signed-in sites
12 valid cookies copied
Cookies only
Passwords stay in your browser
Local transfer
Never written into agent chat
Drive the whole flow
Navigate, fill forms, click controls, and inspect the resulting page state.
Open /checkout
navigation
Enter card
input
Place order
click
Receipt visible
assert
Watch browser actions live
See the page as the agent works and keep screenshots and results in the thread.
Describe what to verify
Give the agent the flow and expected result; it performs the browser actions for you.
“Sign in, finish checkout, verify the receipt.”
checkout.spec.ts · generated
Signed-in sessions
Copy valid cookies from a local browser profile. Passwords are never imported.
A real, headless browser
Tests drive an actual browser — clicks, typing, navigation, and assertions.
Inspect and retry
The agent can inspect the current page state, adjust the implementation, and try the flow again.
The best worktree
management, anywhere.
Every run gets an isolated branch and working tree. See its agent, files, tests, review state, handoff evidence, and cleanup status.
agent/checkout-recovery
Runningagent/session-race
In reviewagent/sidebar-cache
CleanedEvery run stays isolated
Claude → Codex handoff
same worktree · same branch
Plan attached
Transcript attached
Diff · 12 files
Tests · 16 passing
Codex reviewing
Hand off the complete context
1.8 GB reclaimed
Branch preserved
Commits + run history kept
Worktree removed
Reclaim disk without losing work
Integrations
Plugs into the tools you already use
Start from Linear, carry the issue into the run, and review the resulting branch and pull request with GitHub.
Start from a Linear issue
Search connected workspaces and create an agent run from one or more issues.
Attach complete issue context
Bring the issue description and comments into the agent thread.
Let agents update Linear
Agents can read issues, add comments, and update fields through the integration.
Dashboard
Review GitHub pull requests
Inspect the branch diff, checks, reviews, and inline feedback in Zuse.
Manage review feedback
Create, edit, resolve, and delete inline GitHub review comments.
Merge with guardrails
Choose merge, squash, or rebase, and use GitHub auto-merge when allowed.
Deploy agents across every platform
Run agents from desktop today, persistent cloud workspaces in beta, and a mobile companion coming soon.
Agents in your pocket
Coming soon
Follow runs and approve decisions from your phone.

Full control at your desk
Available now
Run agents, inspect changes, and manage worktrees in Zuse Desktop.
Agents that keep running
Live beta
Run isolated cloud workspaces even after your computer goes offline.
Cloud beta
Interested in a hosted workspace?
Register your interest as we open beta capacity to more developers.
Remote access, any machine
Open the Zuse environment already running on another computer from the web or iPhone. Use Zuse Serve for a headless Mac or Linux host.
Zuse CLI
Create the work. Add another session. Keep the context.
Create a fresh worktree from the CLI, start provider sessions in the same chat, and keep the branch and attached context together.
Create a fresh worktree
Run `zuse chat create --workspace fresh` to create the worktree, chat, and first provider session together.
main
Add Codex to the same workspace
Use `zuse session create` for another provider in the same chat and worktree; attach a plan or transcript when needed.
Create → implement → review
Follow each provider session against the same branch and inspect the resulting diff and check evidence in Zuse.
Same worktree
agent/checkout-recovery
Implementation evidence
The chat stays linked to its worktree; plans, transcripts, and files can be attached explicitly.
Independent diff review
Start Codex in the same chat and ask it to inspect Claude Code's branch diff.
GitHub follow-through
Inspect pull request state and checks, or attach failing Actions logs to an agent for diagnosis.
One worktree
Multiple tabs, shared context
Implementation, review, and checks stay side by side on the same branch and working tree.
Build the change in one tab while every file lands in the shared worktree.
agent/checkout-recovery
Implementation
Focused tests
Diff review
Review the same branch and transcript in a second tab without copying context.
Keep typecheck, focused tests, and terminal output beside both agent sessions.
Smart handoff
Use the right model for each stage
Plan, implement, and review with separate provider sessions while carrying forward only the context each one needs.
Plan with one model, continue with another
Start another provider session for implementation and attach the plan, transcript, or files it needs.
Planning session
Claude Code
Plan complete
Implementation session
Codex
Ready to start
Handoff context
Fork when a provider hits its limit
Fork the session to another provider and carry forward a copied transcript or explicit handoff context.
Continue this work
Choose a provider and context
New Codex session
Fork from here
Transcript
Copied
Files
Select to attach
Pick a model per step
Choose which agent handles planning, coding, and review — per task, not per project.
Attach the right context
Add a plan, transcript, diff, or focused files when starting the next session.
You stay in control
Choose the provider and model for each session, then decide what context to carry forward.