Stripe MCP Server

Provides a unified StripeAPI core class that wraps the official Stripe SDK and exposes a consistent interface, with framework-specific adapter layers (LangChain, OpenAI, MCP, CrewAI, Vercel AI SDK, Cloudflare Workers) that translate the core API into each framework's native tool format. Uses a layered architecture pattern where framework integrations inherit from or compose the StripeAgentToolkit base class, enabling code reuse across TypeScript and Python implementations while maintaining framework-native semantics.

Converts Stripe operations into framework-native function schemas (OpenAI function definitions, LangChain StructuredTool with Pydantic models, MCP Tool with JSON schemas) by introspecting the StripeAPI method signatures and generating schema definitions that include parameter validation, descriptions, and type information. Each framework adapter registers tools with its native function-calling mechanism, handling serialization of Stripe response objects back to the framework's expected output format.

Enables agents to monetize specific capabilities by gating them behind Stripe checkout flows. When an agent invokes a paid tool, the toolkit creates a Stripe checkout session and returns a payment link to the user. The agent can then verify payment completion before executing the gated capability. This allows developers to build freemium agent applications where premium features require payment, with Stripe handling the payment processing and checkout UI.

Provides agents with a tool to search Stripe's official documentation using semantic search, allowing agents to look up API details, pricing information, and best practices without leaving the agent context. The toolkit embeds Stripe documentation and uses semantic similarity to retrieve relevant documentation sections based on agent queries. This enables agents to self-serve documentation lookups and understand Stripe capabilities without requiring developers to manually provide documentation context.

Enables agents to work with Stripe connected accounts (platforms with multiple merchant accounts) by accepting account context that specifies which connected account to operate on. The toolkit routes API calls to the specified connected account using Stripe's account header mechanism, allowing agents to manage multiple merchant accounts without requiring separate toolkit instances. This is essential for marketplace and platform applications where a single agent needs to operate across multiple merchant accounts.

Provides identical toolkit functionality in both TypeScript and Python, with framework-specific implementations for each language (TypeScript: LangChain, OpenAI, MCP, Vercel AI SDK, Cloudflare Workers; Python: LangChain, CrewAI, OpenAI). Both implementations share the same core StripeAPI abstraction and expose the same operations, allowing developers to choose their preferred language and framework while maintaining consistent Stripe integration behavior. The toolkit is built on top of official Stripe SDKs (stripe-js for TypeScript, stripe for Python).

Implements the Model Context Protocol (MCP) specification for Stripe operations, exposing all toolkit capabilities as MCP tools that can be discovered and invoked by MCP-compatible clients (Claude, custom agents, etc.). The MCP implementation follows the standard MCP tool format with JSON schemas for input validation and structured output, enabling seamless integration with any MCP-compatible client without framework-specific adapters. Tools are registered with the MCP server at startup and made available to clients through the standard MCP discovery mechanism.

Implements a permission configuration layer that allows developers to selectively enable/disable Stripe operations at toolkit initialization time, controlling which tools are exposed to the agent. The configuration system uses a declarative approach where permissions are specified per operation (e.g., 'create_customer', 'refund_payment') and enforced at the StripeAgentToolkit adapter level before tools are registered with the framework. This prevents agents from accessing sensitive operations like refunds or subscription cancellations unless explicitly permitted.

Provides a complete customer management interface that allows agents to create new Stripe customers with email/name/metadata, list customers with filtering and pagination, update customer details (email, description, metadata), and manage customer metadata for storing application-specific data. Operations are exposed as framework-native tools that handle parameter validation, Stripe API calls, and response serialization. Metadata operations enable agents to attach custom application state to customers without modifying Stripe's core customer schema.

Enables agents to create payment intents with specified amounts, currencies, and payment methods, then track their status through the payment lifecycle (requires_payment_method, requires_confirmation, succeeded, canceled). Agents can retrieve payment intent details, confirm intents to trigger payment processing, and handle payment method attachment. The toolkit abstracts Stripe's payment intent state machine, allowing agents to understand and react to payment status changes without managing the full complexity of Stripe's payment flow.

Provides agents with tools to create recurring subscriptions for customers with specified products/prices and billing intervals, list active subscriptions for a customer, retrieve subscription details, and cancel subscriptions with optional proration handling. The toolkit manages the subscription lifecycle, allowing agents to understand subscription status (active, past_due, canceled, unpaid) and handle common subscription operations without managing Stripe's subscription state machine directly.

Enables agents to create invoices for customers with line items and custom amounts, list invoices with filtering by customer or status, retrieve invoice details, and track invoice payment status. Agents can finalize invoices to trigger payment collection and handle invoice-related operations like sending reminders or marking as paid. The toolkit abstracts Stripe's invoice object model, allowing agents to manage billing workflows without understanding Stripe's invoice state machine.

Provides agents with tools to create refunds for charges with specified amounts and reasons, list refunds for a charge or customer, retrieve refund details, and track refund status through the processing lifecycle. Agents can also retrieve dispute information and understand dispute status (warning_under_review, warning_closed, needs_response, under_review, won, lost). The toolkit abstracts Stripe's refund and dispute workflows, allowing agents to handle payment reversals and customer disputes without managing complex state machines.

Enables agents to retrieve account balance information (available balance, pending balance, reserved balance) and query transaction history with filtering by date range, transaction type, and amount. Agents can understand account financial status and identify specific transactions for reconciliation or reporting purposes. The toolkit abstracts Stripe's balance and transaction APIs, providing agents with financial visibility without requiring deep knowledge of Stripe's accounting model.

Provides agents with tools to record meter events for usage-based billing, allowing customers to be charged based on actual consumption (API calls, storage, compute hours, etc.). Agents can record events with customer_id, meter_id, and quantity, and the toolkit handles event aggregation and billing calculation. This enables flexible, consumption-based pricing models where charges are calculated based on recorded usage rather than fixed subscription amounts.