EightseatGet early access
A product team for bootstrapped founders

Eight seats. One founder.

Eight AI specialists scope, design, build, test, release and market your product. They work on your machine, through your own Claude subscription. You run the table from the web, even from your phone, and approve what ships.

Get early access
npx eightseat

Run once to connect your machine · Free during the private beta

AZA-12Dark mode toggle
Your repo, on your machine
Seat1 · Product ManagerDoingTurning "add dark mode" into stories and acceptance criteria.
The seating chart

Every seat has one job.

Seats are numbered in the order a ticket visits them. Each agent works on its own git branch and leaves you something to read. Five of them stop and wait for you.

Every seat's instructions are a plain text file in your repo. Edit them to fit your project, for example "always use SwiftUI".

How a ticket moves

Ten stops. You're five of them.

Agents move a ticket forward on their own until it reaches a decision only you should make.

    QA fails? Back to the engineer, up to 3 times, then it comes to you with the details.Changes requested? Back to the engineer with your comment.
    Your morning

    Three decisions with coffee.

    Everything waiting on you is in one list, with a plain-English summary from the seat that did the work. Approve from your desk or your phone.

    Needs you3 waiting · 0 blocked
    6

    Merge AZA-12: dark mode toggle

    41 of 41 tests pass on your machine. Risk 1 of 5, nothing blocking. Settings screen and two components changed.

    Reviewer · eightseat/AZA-12-dark-mode

    1

    Scope AZA-15: team invites

    3 stories, 9 acceptance criteria, sized M. One open question about invite expiry.

    Product Manager · PRD saved in your repo

    8

    Launch copy for v1.6

    X thread, App Store "What's New" and a short email. Saved in your repo. Not posted anywhere.

    Marketing · written in your brand voice

    Your data

    We store your board. Never your code.

    Your board lives in your Eightseat account. The work happens on your machine, through a small helper that drives the Claude Code you already pay for. Here is what goes where.

    Your machine

    Eightseat helperAlways on. Runs only your repo's own commands, in folders you approved.
    Claude CodeSigned in with your own subscription.
    Your repoBranches, builds, tests, simulators.

    Code, diffs, logs, prompts and secrets stay here. When you open a log or diff, it streams to your browser and is never stored by us.

    AnthropicReceives prompts from your Claude Code, the same as when you use it yourself.
    Eightseat web appStores your board: tickets, approvals and short summaries from each seat. No code, no diffs, no prompts.
    No markup on AI

    Agents run on your Claude subscription. We charge a flat fee for the board, nothing per token.

    Nothing ships without you

    Agents never push, deploy or post. Merges happen on your machine after you approve, from any device.

    Tested where it runs

    QA uses your real build and test commands, your database and your simulators.

    Pricing

    One ticket. All eight seats.

    Admit one founder
    $9 / month

    Everything included, for every project. Bring your own Claude subscription (Pro or Max).

    • All 8 seats
    • Unlimited projects
    • Editable agent instructions
    • Approve from your phone
    • Full audit log of every run
    • Every update while subscribed
    Questions

    Before you pull up a chair.

    What do I need?

    A Mac, Linux or Windows machine with Node 20+, git, and Claude Code signed in with your Claude subscription. Run npx eightseat once to connect it. Codex CLI and Gemini CLI support is planned.

    Does my code go to your servers?

    No. We store your board: tickets, approvals and short summaries. Code, diffs, logs and prompts stay on your machine, and your Claude Code talks to Anthropic exactly as it does today.

    Do I need to code?

    It helps to be semi-technical: run a command, read a summary, glance at a diff. You don't have to write the code.

    What if an agent gets it wrong?

    QA runs your real tests and sends failures back up to three times. Nothing is merged, released or posted until you approve it, and every run is logged.

    Can I change how the agents work?

    Yes. Each seat's instructions are a text file in your repo. When we improve the defaults, you see the change and decide.

    What if my laptop sleeps?

    The helper keeps it awake while agents work. Close the lid and jobs wait until it's back; the board still works. Want it always on? Run the helper on a Mac mini or a server.

    Your table is ready.

    We're onboarding a small group of bootstrapped founders. Free during the beta, in exchange for a 15-minute call each week.

    Web app on any device · helper for macOS, Linux and Windows