Stripe

Provides a unified StripeAPI core class that wraps the official Stripe SDK and exposes a framework-agnostic interface, with specialized adapter layers (StripeAgentToolkit classes) that translate this core into framework-specific tool formats (LangChain tools, OpenAI functions, MCP resources, CrewAI tools, Vercel AI SDK). The architecture uses a layered pattern where the core handles all Stripe business logic and each framework integration layer only handles format translation, enabling single-source-of-truth maintenance across TypeScript and Python implementations.

Implements a declarative permission model where developers specify which Stripe operations (customer management, payment creation, refund issuance, etc.) are available to agents via configuration objects. The system validates tool invocations against these permissions before executing Stripe API calls, preventing unauthorized operations. Configuration is passed at toolkit initialization and applies uniformly across all framework adapters, enabling fine-grained control over what payment operations an agent can perform without modifying framework-specific code.

Implements a payment gating system where certain Stripe operations (tools) can be restricted to paid customers, with automatic Stripe Checkout integration for payment collection. When an agent attempts to use a paid tool, the system checks customer payment status and initiates a Checkout session if needed. This enables monetization of specific agent capabilities through Stripe Checkout without requiring custom payment logic.

Provides complete abstractions for core Stripe operations including customer CRUD (create, read, update, list), subscription lifecycle management (create, update, cancel, retrieve), invoice operations (create, send, pay, void), dispute handling (retrieve, respond, close), refund processing, balance retrieval, and payment link generation. Each operation is wrapped with proper error handling, parameter validation, and response transformation, enabling agents to perform full payment and billing workflows without direct Stripe SDK knowledge.

Integrates semantic search over Stripe's official documentation, allowing agents to retrieve relevant documentation snippets when they need to understand Stripe API behavior or troubleshoot issues. The system uses embeddings-based retrieval to find documentation sections matching agent queries, enabling agents to self-serve documentation lookups without requiring hardcoded knowledge. This augments agent reasoning by providing real-time access to authoritative Stripe documentation.

Provides a testing and evaluation framework that enables developers to test agent Stripe workflows against synthetic scenarios without hitting production Stripe APIs. The framework includes mock Stripe responses, scenario generators for common billing workflows (subscription creation, invoice payment, refund processing), and assertion utilities for validating agent behavior. Enables safe testing of complex payment workflows and agent decision-making without financial risk.

Provides parallel TypeScript and Python implementations of the Stripe Agent Toolkit with feature parity, allowing developers to use the same Stripe operations (customer management, subscriptions, invoices, disputes, refunds, balance retrieval) in both languages. Both implementations wrap the official Stripe SDKs (stripe-node and stripe-python) and expose identical tool interfaces through their respective framework adapters, enabling teams to build agents in their preferred language without sacrificing capability coverage.

Exposes Stripe operations as MCP resources and tools through a dedicated MCP server implementation, allowing any MCP-compatible client (Claude, custom agents, IDE plugins) to invoke Stripe operations via the standardized MCP protocol. The toolkit implements MCP tool schemas for all Stripe operations and handles MCP request/response serialization, enabling Stripe integration with any tool that speaks MCP without requiring framework-specific code.

Translates Stripe operations into OpenAI-compatible function definitions with proper JSON schema generation, parameter validation, and response handling. The toolkit automatically converts Stripe API parameters into OpenAI function schemas, handles function call invocations from OpenAI models, executes the corresponding Stripe operations, and returns results in OpenAI-compatible format. This enables seamless integration with OpenAI's function calling API without manual schema definition.

Converts Stripe operations into LangChain StructuredTool objects with proper input/output schemas, enabling use with LangChain's agent framework and chat-based agents. The toolkit generates LangChain tool definitions from the core StripeAPI class, handles tool invocation through LangChain's tool calling mechanism, and integrates with LangChain's agent loop for multi-step Stripe workflows. Tools are fully compatible with LangChain's structured chat agents and ReAct patterns.

Provides CrewAI-compatible tool definitions that enable multiple specialized agents to collaborate on Stripe-related tasks through CrewAI's task and role system. The toolkit exposes Stripe operations as CrewAI tools that can be assigned to different agents (e.g., a billing agent, a customer support agent), enabling task decomposition where complex Stripe workflows are broken into subtasks executed by specialized agents. Integrates with CrewAI's memory and collaboration patterns.

Integrates Stripe operations with Vercel's AI SDK through its tool calling interface, including built-in token usage tracking middleware that measures LLM token consumption for each Stripe operation invocation. The toolkit provides Vercel AI-compatible tool definitions and automatically tracks token metrics, enabling cost analysis and usage monitoring for Stripe-integrated AI applications. Middleware transparently captures token data without requiring application-level instrumentation.

Enables Stripe operations to work within Cloudflare Workers environments with built-in OAuth support for Stripe Connect and connected accounts. The toolkit handles OAuth token management, connected account context passing, and Stripe API calls scoped to specific connected accounts. This allows Stripe operations to be executed in serverless Cloudflare Workers with proper authentication and account isolation for multi-tenant scenarios.

Provides abstractions for Stripe's usage-based billing model through meter events, enabling agents to report consumption metrics that trigger billing calculations. The toolkit wraps Stripe's Metering API, allowing agents to emit meter events (e.g., API calls made, tokens consumed, storage used) that Stripe aggregates and uses to calculate usage-based charges. Supports event batching, timestamp handling, and meter configuration management.