Skip to content

MCP setup

Nordva Launch ships an MCP server at https://mcp.nordva.dev that exposes your API as tools any MCP-compatible client can call.

Terminal window
claude mcp add --transport http nordva https://mcp.nordva.dev \
--header "Authorization: Bearer nv_live_..."

The server is registered as nordva in your MCP client. Restart Claude to pick up the new server, then ask:

List my recent changelog entries.

Claude will call list_changelog_entries for you.

Add to ~/.cursor/mcp.json:

{
"mcpServers": {
"nordva": {
"url": "https://mcp.nordva.dev",
"transport": "http",
"headers": {
"Authorization": "Bearer nv_live_..."
}
}
}
}

Reload Cursor → open the MCP panel → confirm nordva is connected.

ToolDescription
setup_checkVerify auth, identify project, report plan and feature flags.
publish_changelog_entryCreate or schedule a changelog entry.
list_changelog_entriesList recent entries (default: 10).
get_insightsOne-call summary of a period against the one before: signups and sources, changelog, feedback, revenue.
get_revenue_summaryMRR, new and churned subscriptions and collected revenue from your connected Stripe account (read-only).
list_contactsPeople across waitlist, subscribers, feedback and Stripe: who is paying, where they came from.
get_attribution_reportWhich signup sources produced paying customers.

The full list of tools with their arguments is at launch.nordva.dev/mcp, with setup pages for VS Code and Claude Desktop.

There are two prompts. plan_integration has the agent read your codebase and write a plan for adding Launch, without changing code: see Plan the integration with your AI agent. weekly_founder_summary composes the read tools into a short weekly write-up.

No tool connects Stripe and none accepts a Stripe key: you connect it once in the dashboard, and the assistant will never ask for a key.

Run setup_check first — it returns the project ID, plan, and any setup warnings before you publish anything.

Tool calls count against your plan’s API rate limit. The MCP endpoint itself additionally allows 120 requests a minute per key and 60 per IP address. If you’re hitting RATE_LIMITED, see rate limits.