Launched this week
Foxpit
Run Claude Code, Codex & OpenCode as one AI team
42 followers
Run Claude Code, Codex & OpenCode as one AI team
42 followers
Foxpit turns your AI coding agents into one team in a 3D office. Fox, the lead, splits the task; Claude Code, Codex and OpenCode build it in real terminals. Shared memory, auto-resume after usage limits, approvals in one place. Windows, macOS, Linux.
Free Options
Launch Team
LoomShow, don't just tell with Loom
Promoted
hey PH 👋 im yigit, i built foxpit alone
the story: one day claude ate 62% of my weekly quota in a single day. i was letting opus write all the code. so i stopped doing that. now claude only plans and reviews, and cheaper models (codex, opencode, whatever's free) write the code
doing that by hand was a mess tho. 5 terminals, copy pasting context, forgetting who's doing what. so i made foxpit
fox (your lead agent) splits the task and hands pieces to the others
every agent runs in a real terminal, you see who's working and who's stuck in a little 3d office
one shared memory in your repo, so the next session doesnt start from zero
when claude hits the 5 hour limit, foxpit waits and continues by itself
your code never leaves your computer, foxpit only checks your plan
its $5 for the first month, then $12. would love to hear what breaks, i fix stuff the same day
Congrats! How does Foxpit decide which agent gets which task? Can I route by task type, like Claude Code for refactors and Codex for tests? And do the agents share one working tree or each get their own branch?
thx! fox (the lead) splits the work and picks who gets what. you can steer it tho, every teammate has its own agent + model in team settings, so u can literally say “have codex write the tests” and it does. each agent can get its own git worktree on its own branch so they dont step on each other
I've hit that same Opus quota wall — Claude eating most of a week's quota in a day is exactly why I started keeping Claude on planning and review and letting cheaper models do the bulk of the writing. Doing that by hand across terminals is the mess you describe, so having one lead split the work and keep the shared memory in the repo is the part that appeals to me. The auto-resume after the 5-hour limit is the other one — my longer builds always seem to die mid-task. Curious how Fox decides a cheaper model's output is good enough to pass on versus sending it back for another round.
@jcbuilds yeah thats exactly how i started using it lol, claude plans + reviews, cheaper models write. when a teammate finishes fox gets a delivery card with changed files + test results, reads the diff, and if its off it sends it back to the same agent. and on limits, the agent parks and continues on its own when the limit resets, or u keep the same chat on another account
Huge congrats on the launch, Yigit!
I think every solo builder has hit that Opus quota wall and had to just stop working for the day, so this is a massive pain point you are solving. The 'little 3d office' to visualize the terminal states is a brilliant UX touch, too.
Quick question on the workflow: how does Fox (the lead agent) actually verify that the code written by the cheaper models is correct? Does it run local tests and loop the cheaper models back in if it fails, or is it strictly doing a syntax/logic review?
Wishing you a great launch day
@muhammad_salman39 thank u man! fox looks at the actual diff and the test results from the delivery card, not just vibes. if tests fail or the diff is wrong it sends it back for another round
Coordinating Claude Code, Codex, and OpenCode with shared memory is a useful approach. How does Foxpit prevent conflicts when multiple agents modify the same files, and can it automatically resolve those conflicts before merging?
@mahmoud_shawkat mostly by not letting it happen, each agent works in its own worktree so they never touch the same checkout. merging back is a reviewed step, u see the diff first. no magic auto-resolve tho, if two agents changed the same lines fox or u decide which one stays