LaunchRepo.

Works with

Launch your SaaS on directories with Cursor

Cursor’s agent reads your workspace and can use MCP servers — which is all LaunchRepo needs. Open the cloned repository, add the browser server, paste one prompt.

Cursor users get a slightly different experience from a terminal agent: the playbooks, product.json and submissions.csv are open in the editor next to the agent panel, so you can read what the agent reads and correct a fact before it reaches a directory. The setup is the same three things — the repository, Python 3.10+, an MCP browser server — and the workflow is the one described on the how-it-works page.

What Cursor needs

The repository, opened as a workspace
Accept the GitHub invitation after purchase, clone the private repository and open the folder in Cursor. The agent reads AGENTS.md, the runbook and the playbooks from there.
Python 3.10+ on macOS or Linux
The workspace script, record.py and the local dashboard use the Python standard library only. Cursor’s integrated terminal runs them; Windows is not supported.
An MCP browser server in Cursor’s MCP settings
Cursor’s agent reaches Chrome through an MCP server. Register Chrome DevTools MCP as described in BROWSER-SETUP.md in Cursor’s MCP configuration; follow Cursor’s own settings UI for the exact place.
Permissions in authorizations.json
Free registrations, final submissions, mail actions and badges are separately scoped grants recorded once per product during onboarding. The agent respects them on every resume.

Connecting Cursor to your Chrome

The rule in BROWSER-SETUP.md is the same for every agent: the browser you can see, one long-lived connection per batch, the tracker never touching Chrome itself. If Cursor already has a working browser integration, keep it; do not run two.

Preferred route: enable remote debugging in your everyday Chrome (144 or newer, via the inspect page) and add Chrome DevTools MCP with auto-connect to Cursor’s MCP servers. Ask the agent for a harmless browser action, approve Chrome’s prompt in the intended window and confirm the visible tab moved.

Fallback: a dedicated Chrome with its own user-data directory and a local debugging port; sign in to your directory and mail accounts there and point the MCP server at that endpoint instead. Never both modes, never a public port, never someone else’s endpoint.

Then the smoke check: a blank tab, a public directory homepage, the heading, which tab moved, and a signed-in directory whose account menu shows the right identity. Record pass/fail only.

The start prompt

The public start prompt from START.md, unchanged. Paste it into Cursor’s agent with the cloned folder open: it creates your private workspace next to the checkout, starts the dashboard and answers with a 127.0.0.1 link.

START.md — paste as-is, in English
Read AGENTS.md, then follow "Start the dashboard for the owner". Create my private
workspace if I do not have one yet, start the dashboard, and give me the link.

The playbooks and instructions are English Markdown, which is what coding agents handle best. Your product copy can be in any language a directory accepts; the agent answers you in your language.

Onboarding continues in the dashboard — product brief, identity, permissions, browser smoke check — and each later prompt (first launch, resume, review, handoff) is there to copy back into Cursor. The files it writes are the ones open in your editor.

What a run looks like

The illustration below is the one from the home page: a small batch, each platform ending in the state the tracker observed. Real statuses come from your submissions.csv; nothing here measures Cursor.

What still needs you

LaunchRepo is deliberately not a background service. These moments stop the run and are recorded as human handoffs; the agent never works around them:

CAPTCHAs and bot checks
The agent stops and records prepared_needs_human; the dashboard lists the platform under “needs you”. You solve it in the same Chrome window.
2FA and verification codes
Codes arrive in the mailbox signed in inside the connected browser. Forwarding a code by hand is a supported fallback and is recorded as a human handoff.
Paid upsells
A paid-only route becomes deferred_paid with a note. The agent never buys, never starts a card-backed trial and never accepts an unrelated offer.
Confirming “live”
A platform is marked live only with a public URL and your confirmation that you reviewed it. Drafts, receipts and thank-you pages keep their own states.

No benchmark, no ranking: we claim compatibility for Cursor because it meets the two requirements above, not because we measured it against other agents.

Questions

Do I use Cursor’s chat or its agent mode?

The agent needs to run scripts and call the MCP browser server, so use the mode in your Cursor version that can do both. We do not document Cursor’s UI; its own settings describe where MCP servers and tool permissions live.

Can I edit product.json in the editor instead of the dashboard?

Yes. The dashboard and the agent write the same files; editing them in Cursor is fine. The agent rereads product files after the dashboard or you have changed them.

Does Cursor need a paid plan?

We make no claim about Cursor’s plans. Agent usage is billed by your provider and is not part of the LaunchRepo licence.

Does the browser integration see my other tabs?

A CDP or DevTools MCP connection controls the connected browser, including authenticated pages. BROWSER-SETUP.md tells you to close unrelated sensitive tabs and use only the product’s permitted sites.

Can I switch to Claude Code or Codex later?

Yes. The workspace, tracker and permissions are plain files; any agent that reads files and drives a browser continues where the last one stopped. The licence covers unlimited agents.

One payment, every product, any agent

The launch price is €99.98, once. It rises 20 % after every 20 sales; the pricing page shows the live price, what is in the repository and the licence in short.

See pricing