Stripe Founder

npm version License: MIT Node.js

The Stripe co-pilot for indie SaaS founders. Ask your AI about MRR, churn, failed payments, and at-risk customers in plain English — and get answers shaped around the questions a founder actually asks, not raw API dumps.

Prefer a managed hosted version? Install in one click from mcpize.com/mcp/stripe-founder. Free tier includes 200 requests per month.

The problem

Stripe's API is exhaustive. The Dashboard answers most questions but requires clicking through ten screens. Existing Stripe MCPs are thin CRUD wrappers — they let an AI list charges or retrieve a customer object, but they don't answer founder questions.

This MCP is opinionated. Each tool maps to a question a founder wakes up asking: "What's my MRR?", "Who's about to churn?", "Is dunning recovering my failed payments?" The math is done for you. Tools return structured data shaped for an LLM to summarize naturally — headline number first, supporting context after, raw line items last.

What it looks like in practice

You're in Claude Code or Cursor. You ask a question, the AI calls the right tool, and answers in plain English:

You: What's my MRR right now and how does it compare to last month?

AI: Your MRR is $4,820, up 12.3% from $4,292 last month. You have 47 active
    subscriptions. There are also 3 trials in progress worth an additional
    $147/mo if they convert.

You: Who's at risk of churning this week?

AI: 4 customers, $680 MRR at risk. Top concern is acme-corp.com ($290/mo,
    3 failed payment attempts, last retry failed 6 hours ago). Two others
    are past-due, one has a payment method that just expired.

You: How well is dunning recovering failed payments this month?

AI: 50% recovery rate on 14 failed payments — but only 4 have reached a
    final outcome. 2 recovered ($310 in revenue), 2 final failures, 10
    still in retry. Top unrecovered: $290 from acme-corp.com.

No SQL, no dashboard clicking, no manual API calls.

Install

You have two options. Pick whichever fits your workflow.

Option A — Hosted (recommended)

The fastest way to get started. No Node setup, no self-hosting, no infrastructure.

Install via MCPize →

You provide your Stripe key, MCPize handles the rest. Free tier includes 200 requests per month with no expiration. Pro tier is $14/month for 5,000 requests with predictable overage pricing.

Option B — Self-hosted

Run it yourself, free forever, full control.

Requires Node 20 or later.

1. Create a Stripe restricted key with read-only access at dashboard.stripe.com/apikeys. The MCP needs read access to your billing data: customers, subscriptions, charges, invoices, payment intents, products, prices, and disputes.

2. Add to your MCP client config:

Claude Desktop

Config file: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows)

{
  "mcpServers": {
    "stripe-founder": {
      "command": "npx",
      "args": ["-y", "stripe-founder-mcp"],
      "env": {
        "STRIPE_API_KEY": "rk_live_..."
      }
    }
  }
}

Cursor

Config file: ~/.cursor/mcp.json

{
  "mcpServers": {
    "stripe-founder": {
      "command": "npx",
      "args": ["-y", "stripe-founder-mcp"],
      "env": {
        "STRIPE_API_KEY": "rk_live_..."
      }
    }
  }
}

Claude Code

claude mcp add stripe-founder -- npx -y stripe-founder-mcp

Then set STRIPE_API_KEY=rk_live_... in your shell environment or project .env file.

Development from source

git clone https://github.com/PrimeEcto/stripe-founder-mcp.git
cd stripe-founder-mcp
npm install
cp .env.example .env   # then fill in your STRIPE_API_KEY
npm run dev

Tools

Tool	What it answers
get_mrr	Current MRR with delta vs a comparison period, plus trialing subscriptions tracked separately as potential MRR
get_growth_metrics	New MRR, churned MRR, expansion, net new MRR, and growth rate for a period
get_churn_summary	Churn count, voluntary vs involuntary, gross/net churn rate, top cancellation reasons
get_failed_payment_recovery_rate	How well dunning is working — recovery rate, recovered revenue, top unrecovered failures
list_at_risk_customers	Customers at risk of churning due to past-due status or repeated payment failures
list_recent_signups	New customers in a date range, with MRR contribution and payment method status
list_disputes	Open chargebacks with evidence deadlines, or filter to won/lost
get_customer_profile	Full picture for one customer: LTV, plan, payment status, and recent timeline
get_support_context	Everything needed to handle a support ticket: risk flags, refund-eligible charges, prioritized timeline

All time-based tools accept human-readable date ranges: "last_30_days", "this_month", "2026-Q1", "March 2026", or explicit ISO ranges.

get_customer_profile and get_support_context accept email, name, or Stripe customer ID — no need to look up the ID first.

What this is not

• Not a write tool. It can't create products, issue refunds, or cancel subscriptions. Read-only by design.
• Not Stripe Sigma. No SQL queries — these are pre-built, opinionated answers to common questions.
• Not a Dashboard replacement. Use it alongside the Stripe Dashboard, not instead of it.

Safety

• Read-only enforced at runtime. A guard blocks any non-GET Stripe request at the HTTP client level. The codebase has zero write paths.
• Restricted keys recommended. Use rk_live_... scoped to read access only, not your full secret key.
• Test mode auto-detected. If your key starts with sk_test_ or rk_test_, every tool response includes stripe_mode: "test" so you never confuse environments.
• No telemetry. No data leaves your machine beyond the calls to Stripe's API. No analytics, no external logging.

Contributing

Issues and pull requests welcome. If you find a bug, open an issue with the tool name, the input you passed, and what you expected vs what you got. If you want to add a tool, read AGENTS.md and SPEC.md first — the project has strong opinions on what belongs here.

License

MIT