Choose the work
Start with a prompt or Jira ticket. Choose your folders, model and reasoning level.
FLEET EARLY ACCESSRun in parallel.
Run in parallel.
Stay in control.
Run Claude Code and Codex across your repositories. Fleet keeps the work in separate worktrees, opens PRs and follows up on failed checks and review comments. You make the merge decision.
Explore the demo ↗ EXPLORE FLEETInteractive demo · Sample data
Your work, all in one place.
Explore a ticket, pick up a thread, and follow the PR.
Work on @ORB-142: Keep keyboard focus inside the command menu
You
Work on @ORB-142: Keep keyboard focus inside the command menu
◎
Fleet · Verification passed. PR #82 is open; Fleet is watching CI and review feedback.
Codex · Orbit
The keyboard navigation change is verified and published. I added coverage for Tab navigation and returning focus to the trigger.
8 tool calls Orbit
- Read project context and repository instructions
- Inspect keyboard navigation and regression coverage
PR #82 · Checks passed
◎ Waiting
Fleet is watching PR #82 for CI and review feedback.
8 tool calls
Checks passed · Not merged
Orbit / website
Execution
GPT-6 Astra · High
Independent worktree
Execution settings are fixed for this sample thread.
PR workflowFleet creates a PR after verification and watches CI and feedback. You review and merge.
@orbit ORB-142
YOUR TOOLS, WORKING TOGETHER✳ Claude Code⌘ Codex⑂ Your repositories◈ Your workflow
Product preview · illustrative workspace with synthetic data
FROM FIRST PROMPT TO YOUR FINAL REVIEWLess juggling.
Less juggling.
More code ready for review.
One thread carries the task through implementation, verification and PR follow-up. Your agents can finish a session without losing the work.
Run in parallel
Give repository contributions their own worktrees. Run Claude Code and Codex threads side by side.
Let Fleet follow through
Fleet verifies changes, opens PRs and starts follow-up sessions for CI failures and review feedback.
Make the final call
Get notified when the required PRs are ready. Review and merge on GitHub, then confirm acceptance.
A WORKSPACE THAT HOLDS THE CONTEXTYou steer.
You steer.
Fleet handles the follow-up.
From the first change to the last review comment, the context stays with the thread.
◈ Orbit1 needs attention
◇ Developer tools2 working
◇ Infrastructure1 waiting
More projects.
Less context switching.
Keep repositories, threads and project defaults together. Return to the work and its decisions, even after an agent session ends.
⑂ orbit-web · PR #83Ready
⑂ orbit-infra · PR #84Waiting
↳ Land after orbit-web#83
One request.
Every repository it takes.
Bring the frontend, API and infrastructure into one thread. Each contribution keeps its own worktree and PR; Fleet tracks the dependencies between them.
Feedback ↳
Fix + verify ✓
Your review ◈
The session ends. Fleet keeps watching.
Keep PR follow-up
from becoming a job.
While the app is open, Fleet watches checks and reviewer feedback, starts focused repair sessions and verifies the fixes. Get notified when the work is ready for your review and merge.
More ways to keep work moving
Keep your next instruction ready
Queue messages while an agent works. Send a follow-up to all unfinished contributions or focus on one repository.
Bring the right context
Tag reference folders in your prompt. Fleet keeps read-only context separate from repositories with their own build agents.
Start where the work lives
Browse Jira tickets and launch a thread with the description and comments attached. Or start directly from a prompt.
Return without starting over
Search, pin and organize persistent threads by project. Saved handoffs carry the work into the next agent session.
BEYOND A WALL OF TERMINALSThe agent finished.
The agent finished.
Fleet keeps following through.
Stop carrying review comments and failed CI logs between tabs. Fleet brings new feedback back to the agent with the task’s saved context.
See the workspace in action ↗A place for every change
Projects, threads and isolated repository contributions.
A signal when you’re needed
Questions, blockers and a notification when the work is ready for review.
Context that outlasts a session
Conversation history, handoffs and PR watching.
BUILT AROUND REAL REPOSITORIESDifferent hats.
Different hats.
The same need for control.
For solo builders, technical leads and the people who keep software working.
Ship the backlog
Run a feature, a bug fix and a refactor in separate worktrees.
Turn findings into fixes
Reproduce a regression, add a test and review the patch.
Review infrastructure changes
Update CI and deployment configuration with repository checks.
Remediate with context
Review a reported finding and supervise a targeted code change.
BUILT FOR THE WHOLE CHANGEMore connected work.
More connected work.
Still your call.
One feature can touch several repositories. Fleet keeps their progress and landing order connected in the same thread.
One thread. Multiple repositories.
Bring up to eight folders or repositories into one thread. Each repository gets its own agent workspace, verification and PR.
Keep the whole change moving.
Fleet tracks landing dependencies and checks for conflicts between contributions before calling the work ready.
Know when it’s your turn.
Get a notification when all required PRs are ready for human review. Review and merge, then confirm acceptance to complete the thread.
What happens after an agent finishes? +
Fleet keeps the conversation, session history and handoffs together. While the app is open, it watches PRs after agents stop and starts focused follow-up sessions for new CI failures, review feedback and conflicts.
KNOW WHAT YOU’RE STARTINGYour workspace.
Your workspace.
Clear boundaries.
Thread history and project settings stay local. Your coding agent uses its provider; connected PR and ticket workflows use GitHub and Jira.
Bring alongA local folder, or Git + GitHub for PR workflowsClaude Code or Codex, installed and authenticatedYour own agent subscription or usage billing
A few good questions.
Is the $49 pilot still available?
No. The earlier paid pilot offer is paused while we test whether Fleet solves a workflow that existing tools miss. The current Windows evaluation has no Fleet charge. Bring your own Claude Code or Codex subscription or API billing.
When can I try Fleet?
We’re inviting a small number of Windows developers into free workflow evaluations. Request one and we’ll email you about fit and setup. There’s no announced general launch date yet.
What am I signing up for?
A conversation about whether Fleet helps with your Windows development workflow. We store your email address, signup time and where you joined on this page in Supabase. Requesting an evaluation does not create an account or charge you.
What will I need to use Fleet?
An installed, authenticated Claude Code or Codex CLI and a local folder. Add Git and connect GitHub for isolated repository work and PR follow-up. You bring your own agent subscription or usage billing.
Can I use Claude Code and Codex together?
Yes. Run Claude Code and Codex threads in the same workspace. Choose the model and reasoning level when starting work; each CLI uses its own authentication and provider account limits.
Do I need a ticket tracker?
No. Start with a prompt and a local folder or repository. Connect Jira Cloud to browse tickets and attach their context to a thread. Linear and Notion connections are not available in the current app.
Can one launch change multiple repositories?
Yes. A thread can contain contributions across repositories, each with its own worktree, verification and PR. Fleet tracks their dependencies and waits for all required contributions before marking the thread ready for human review.
How much control do I keep?
Choose the work, answer questions and pause or steer agent sessions. Fleet verifies changes, opens PRs and follows up on CI failures and review feedback while the app is open. It notifies you when the required contributions are ready for human review. You review and merge on GitHub, then confirm acceptance in Fleet.
Where does my data go?
Fleet saves thread history and project settings locally. Git repository work can run in isolated worktrees; ordinary folder work edits that folder directly. Your coding agent sends context to its provider, and connected GitHub and Jira workflows use those services. Cloud sync is not connected in the current app.
Your next few tickets.
Already moving toward review.
We’re looking for developers with a real workflow that their current agent tools handle poorly. Tell us what you use today; if Fleet might help, we’ll set up a free Windows evaluation on your work. Your agent subscription or API usage is separate. Paid plans are paused while we verify product fit.
Explore the demo ↗