Foxl Code
A fleet of autonomous coding agents, available at code.foxl.ai and inside the Foxl app
Foxl Code
Foxl Code is a fleet of autonomous coding agents. You describe work in natural language; an orchestrator plans it, spawns coding agents - on Foxl's free Remote machine, on your own computer, or on a box you have added - and drives them all the way to open pull requests - opening GitHub Issues, pushing branches, and pausing for your review when it needs you.
Foxl Code is available in two places:
- code.foxl.ai - the standalone web interface. Same bundle as the app, pinned to this one product, so it has no product switcher.
- Inside the Foxl app (web at app.foxl.ai, desktop on macOS/Windows/Linux, and the iOS and Android apps) - pick Code from the switcher at the top of the sidebar. Same account.
Foxl Code is in private beta, and access is granted per account. Sign in, and an account not yet on the list sees a waitlist screen in place of the product; it says the account is on the waitlist and will be notified when access is granted. Until then Settings hides the Code panes rather than offering forms that cannot save. The rest of the Foxl app is unaffected. Questions about access go to [email protected].
By default Foxl Code runs on Amazon Bedrock AgentCore: each coding agent gets its own isolated microVM with a real filesystem and full terminal access - it is not a sandboxed snippet runner. A task can instead run on your own computer, or on a machine you add under Settings > Code > Machines. See On your own computer.
The short version
Foxl does not sell model access, so before the first task you bring your own: a Claude Pro/Max or ChatGPT Plus/Pro subscription synced from the Foxl desktop app, or an API key. The setup checklist has a step for it that ticks itself off once one is in place. Then:
- Chat with the orchestrator about what you want built or fixed.
- For a small request it starts a coding agent in the same turn and tells you what it started.
- For work that fans out across several pull requests it writes a Task Document - a plan where each checkbox becomes one pull request - and waits. You confirm once, and it spawns a coding agent per checkbox.
- Each agent runs a real coding CLI - Claude Code, Codex, Kiro, Cursor, Hermes or OpenCode - on your own subscription or API key, pushes a branch, and opens a PR.
- If an agent needs you, it pauses in review; you reply and it resumes from exactly where it stopped.
Read next
- How it works - the orchestrator, Task Documents, and the plan-confirm-fan-out flow.
- Tasks - the task lifecycle and what each state means.
- Work - what needs you, what is running, what shipped, across every repository.
- Coding agents - the six backends, what each one runs on, and how to pick one.
- On your own computer - run a task on your machine with your own CLI login and GitHub credential.
- GitHub - connecting repos, the GitHub App, branches, and PRs.
- Budgets and tiers - the credit model, per-task budgets, and plan limits.
- Autonomous loop - let the orchestrator check back and drive tasks on its own.
- Resume from review - replying to a paused agent and continuing its session.
- Review on approval - approve a PR and let an independent reviewer decide whether it merges.