// the find
stripe/ai
One-stop shop for building AI-powered products and businesses with Stripe.
Stripe's repo of SDKs and tooling for billing usage from LLM applications. @stripe/ai-sdk connects Stripe billing to Vercel's ai package, and @stripe/token-meter wraps the native OpenAI, Anthropic, and Gemini clients without a framework dependency. It is aimed at developers who already use Stripe and want to meter model usage per customer.
token-meter sits directly on the provider SDKs, so a team that doesn't use Vercel's AI SDK can still meter usage without adopting a framework. The hosted MCP server at mcp.stripe.com authenticates over OAuth, so no secret key has to live in an agent's config file. The benchmarks directory includes graded checkout tasks with Python and Ruby graders, which is more evidence than most agent-tooling repos offer for their own claims.
The README is mostly install commands and links. The SDK APIs live under llm/ai-sdk and llm/token-meter, which are cut off in the visible tree, so nothing here shows how metering works or how streamed and tool-call tokens are counted. Skills installed with `npx skills add` do not auto-update, which the README admits, so teams on that path will drift from current docs. Beyond the MIT-licensed code, the value depends on Stripe's hosted MCP service and on a Stripe account.