The multiplayer AI coding workspace.

Your whole team drives live agent sessions together, from a browser or your terminal.

Claude Code and Codex made one person faster, alone.Coshell makes the whole team faster, together.

Why now

The agent runs longer than you can watch.

Agents now run tasks that stretch across hours and your whole codebase. Watching that alone, in a window only you can see, is how context gets lost and mistakes slip through.

Same live session
Everyone works in a shared session, not parallel copies.
Repo loaded once
The agent reads the whole repo once, for the whole team.
Browser or terminal
Join the same session from a browser or your terminal.

What “multiplayer” actually means.

Not a transcript you share after. The live session you’re in together.

01 · live, not later

Drop into the work already in motion.

You and your teammates work in one session, live, watching the same work land on the same running environment. Closer to a Google Doc than to git: one shared source of truth, not copies you merge back later.

02 · when edits overlap

Fast parallel work, nothing clobbered.

One shared environment means the worst collisions never form: there is no diverging copy to merge. But two agents can still reach for the same lines of the same file, and the engine underneath is last-write-wins. Coshell catches that and steers the second edit before it overwrites the first.

03 · one context, one bill

Never re-explain the codebase.

A shared, always-warm session ingests your context once and keeps it cached, instead of every developer paying to reload it. Cached reads cost a fraction of fresh input, and the more context-heavy the work, the more that compounds.

Up and running in under a minute.

01

Open a drive

Run it on your own machine with the CLI, or spin up a Coshell-hosted one. Your repo, terminal, and browser.

02

Bring your team in

They join from the browser or the CLI. Nothing to install, nothing to migrate.

03

Prompt one agent, together

Everyone is in the same live session. Prompt it, watch it work, redirect it, hand it off.

Runs on your terms

Your own keys

70+ models, scoped per drive. Admins pick the allowed set and a fallback. Usage is billed by your provider, not by us.

Your own privacy

Each session is private to you, shared with named teammates, or open to the whole drive. Your call, per session.

Your own machine

Self-host a drive on your own hardware and your code never leaves it. It pushes to your real git repos and runs on your compute.

works withand 70+ more

Free on your own machine.

Run a drive on your own hardware and bring your team into the same live session. Free for up to five teammates. Past that, we’ll set your team up.

Free
$0
  • A self-hosted drive on your own machine
  • Up to 5 teammates in your org
  • Your compute, unmetered
  • Free models, or bring your own keys
  • Browser and terminal, with an integrated localhost browser
Start free
Teams
Let’s talk
  • Everything in Free
  • More than 5 teammates
  • Coshell-hosted drives, if you’d rather not self-host
  • Priority support and onboarding
Email us

A self-hosted drive runs on your hardware, so your code and compute stay yours.

Stay in your terminal.

Install the CLI, then attach to any drive from your shell. You are in the same live session as everyone on it: the same agent, the same teammates, the same running environment.

$ curl -fsSL https://coshell.ai/install | bash

Built on OpenCode, compatible through v1.17.10.

Questions, answered

What does multiplayer actually mean here?

Several people in the same live session at once, working with the agent and watching its work as it happens, on a shared filesystem and terminal. It’s closer to Figma or a Google Doc than to git: one source of truth everyone sees live, instead of each person taking a copy and merging later.

Isn’t this just a read-only share link?

No. A share link is a read-only transcript of work that already happened. Coshell is the live session you’re in together: presence, follow-along, one prompt stream, the same running environment. Not a recording you send after. The session you’re in, now.

Won’t two people on one agent just clobber each other?

No, for two reasons. Structurally, one shared environment means there’s no diverging copy to lose so the worst class of collision can’t form. And when two agents reach for the same lines of the same file, an overlap layer catches it and holds the second edit before it clobbers, steering that agent to take other lines, wait, or let the first land first.

Can I work privately, or is everything shared?

Every session has its own privacy control. Keep a session private to just you, share it with a few specific teammates, or open it to everyone on the drive. Experiment on your own, and open a session up only when you’re ready.

Do we have to be online at the same time?

No. A drive is a persistent, warm session, so multiplayer is real-time when you want it and a clean handoff when you don’t. A teammate in another timezone opens the same drive and picks up where you left off, with the agent’s context intact, instead of cold-starting.

Why won’t Anthropic or GitHub just build this?

A model owner is built to make one agent better. The hard part for a team is coordination across people and across models, and that layer is only useful if it stays neutral: many models on your own keys, your own git, your own infrastructure. A model owner is disincentivized from building exactly that, because neutrality works against their own model.

Where does our code live? Can we keep it in-house?

Work in a drive pushes to your repositories as a normal git operation, so your branching, review, and CI stay as they are, and you’re never locked in. Self-host the drive on your own machine and your code never leaves it. See the security details.

Get your team in the same session.

Open a drive in the browser, or attach from the CLI. Your whole team works in shared live sessions, on the same codebase.