wtf is toolport-mcp-servers?

toolport-mcp-servers turns a company's public API description, called an OpenAPI spec, into a full set of tools an AI assistant can call directly. It currently ships ready made servers for three services: Stripe with 587 tools, Vercel with 333 tools, and Clerk with 224 tools. These are unofficial, generated from each vendor's published API documentation, and are not affiliated with Stripe, Vercel, or Clerk. The project uses the Model Context Protocol, or MCP, a standard way for AI tools like Claude Desktop or Cursor to connect to external services. You point your AI client at one of these servers with a single npx command and an API key for that service, and the assistant gets access to the entire API as individual callable actions, such as create_customer or list_organizations, instead of a handful of generic tools. The author explains that most automatically generated API tool sets end up with confusing names like PostCustomers or GetUsersUserIdOrganizationMemberships, which makes it hard for an AI model to figure out which tool to use for a given request. This project renames those tools into clearer, intent based names and writes richer descriptions using the actual vocabulary of each domain, such as charge, refund, and dispute for payments. The author reports this improves how often the right tool gets found in testing, citing a jump from 91 percent to 96 percent success on a benchmark of Stripe tools. Because exposing hundreds of tools at once can overwhelm an AI agent's context, the README also mentions a separate paid product called Toolport that can sit in front of these servers to let the agent search for tools on demand instead of loading all of them, and to require human approval before destructive actions like refunding a charge or deleting a deployment. The project also documents how to add new vendors by writing a small configuration file and running a prep script.