multi

Close your laptop. Your agents keep shipping.

Claude Code, Codex and Copilot, each in its own cloud sandbox. Alerts, issues and failed builds start them, and they come back as pull requests.

Start freeTalk to us

From a connected repo to a merged pull request.

Set it up once. After that, work starts on its own and you only step in to review.

  1. 1

    Group your repos into a project.

    Multi builds it once per commit and runs your setup script, so every session starts with dependencies installed.

  2. 2

    Pick what starts the work.

    Turn on a built in agent, start from a template, or write your own prompt and attach any trigger.

  3. 3

    Agents work in parallel.

    Each session gets its own sandbox. Watch any of them live, answer a question, or open a terminal.

  4. 4

    Review and merge.

    Finished work lands as a pull request from an agent/ branch. Merge it, or ask for changes in the same session.

Watch it work. Step in anytime.

Connect a tool. Its agents are ready.

Connecting GitHub, Sentry, Linear or Slack creates these agents with their triggers. Start from 20 more templates, or write your own.

Triggers from

  • GitHub
  • GitLab
  • Slack
  • Linear
  • Jira
  • Sentry
  • PagerDuty
  • Zulip
  • Schedule
  • Webhook

  • PR review

    Pull request ready

    Reviews each pull request when it's ready, and again after every push.

  • Issue fixer

    Issue opened

    Fixes each new issue it can and opens a pull request, or says why not.

  • CI fixer

    Workflow failed

    Fixes a failed CI run on your branches and opens a pull request with the fix.

  • Sentry fix

    Alert triggered

    Fixes the issues an alert rule sends, and reports back on the issue.

  • Linear agent

    Issue delegated

    Works on issues delegated to it and answers follow up questions in the session.

  • Slack @mention

    @multi in a thread

    Does what a message addressed to it asks, and replies in the thread.

Agents get a sandbox and a branch. You keep main.

  • Pushes only to agent/ branches.

    Agents open pull requests. They can't merge, tag or push anywhere else, and you decide what lands.

  • A sandbox per session.

    Each session runs in its own container with your repos checked out. Your setup script never sees your credentials.

Start free. Pay for more sandboxes.

Every plan includes monthly credit for the machines your sessions run on. Model usage stays on your own provider account.

  • Free

    $0/ month

    For trying it out

    Try Multi on your own.

    • $5 of compute credit each month
    • 1 member
    • 5 sandboxes at once
    • 5 projects
    • 1 vCPU machines, 10 GB disk
    Start free

  • Pro

    $60/ month

    $600 billed yearly

    For individuals and small teams who need consistent usage.

    • $60 of compute credit each month
    • Up to 5 members
    • 10 sandboxes at once
    • Unlimited projects
    • Up to 4 vCPU machines, 50 GB disk
    Choose Pro

  • Max

    Recommended

    $200/ month

    $2000 billed yearly

    For teams with higher usage and more concurrent tasks.

    • $200 of compute credit each month
    • Up to 10 members
    • 30 sandboxes at once
    • Unlimited projects
    • Up to 8 vCPU machines, 100 GB disk
    Choose Max

  • Enterprise

    Custom

    Built around your team

    For organizations with their own security and infrastructure needs. Tell us what you need and we'll plan it with you.

    • Everything in Max
    • SSO and SAML login · coming soon
    • Sandboxes on your own machines · coming soon
    • Custom limits on members and sandboxes · coming soon
    Let's talk

Questions

Ask us directly
What is Multi?

A cloud platform for coding agents. Each session runs Claude Code, Codex or Copilot in its own sandbox with your repositories checked out, and you follow it from the Multi app on the web, desktop or your phone.

What starts a session?

You, from the Multi app, or a trigger: a GitHub or GitLab event, a Slack, Linear, Jira, Sentry, PagerDuty or Zulip event, a schedule, or a webhook.

Can agents push to my main branch?

No. Agents push only to branches under agent/ and open pull requests. They can't merge, tag or push anywhere else.

What do credits pay for?

The machines your sessions run on, billed by the hour by size. Each plan includes a monthly allowance. Agents use your own model provider credentials, and Multi doesn't charge for model tokens.

How does my repo get set up?

Group repositories into a project. Multi builds it once per commit, running your setup script if you commit one, so sessions start ready.

What happens when credit runs out?

New work is refused and running sandboxes stop until credit is added, so usage never runs past your balance.

Hand off the next task.

Connect a repo, pick an agent, describe the work.

Start a sessionRead the docs