A fleet of Claude Code, controlled by chat

One prompt.
A fleet of agents.

Tell Foxl Code what you want. It spins up isolated Claude Code agents in parallel and lands pull requests when they're done - on the web, desktop, and your phone.

No credit card required · Bring your own Anthropic, AWS, or OpenAI key

How it works

From paste to pull request,
in one click.

Step 1

Connect a repository

Install the Foxl Code GitHub App once and pick the repos you want the agent to operate on. Permissions are scoped per install.

Step 2

Describe a task

Paste a rough TODO or link a GitHub Issue. Foxl drafts a structured Issue, kicks off a Claude Code session in an AgentCore micro-VM, and starts streaming.

Step 3

Review the PR

The agent commits, pushes a branch, and opens a draft pull request. You review like any normal PR. Stop the run anytime; the agent commits before shutting down.

What you get

A development partner that ships.

Real isolation

Every task runs in a fresh AgentCore micro-VM with a per-task GitHub installation token. No state leaks between tasks.

Hard budget caps

Set a per-task or account-wide cap. The agent commits and stops cleanly when it hits the limit. No surprise bills.

Bring your own keys

Use your Anthropic, AWS, or OpenAI keys. Foxl handles orchestration and budgets; spend stays on your billing.

Live streaming

Every shell command, file edit, and model message is fanned out over SSE so you can watch the agent work in real time.

Issue-first by design

Foxl drafts the GitHub Issue from your paste, then implements it. Every task lands as a regular Issue + draft PR you can review.

On every device

Use Foxl Code at code.foxl.ai or inside the Foxl app - web, desktop, and phone. One account shared with Foxl Desktop and Notes, and your tasks and their history follow you between them.

Ready to ship?

Sign in with GitHub, Google, or Apple. Your first task takes under two minutes - at code.foxl.ai or inside the Foxl app.