Paddle Docs MCP Server

The Paddle Docs MCP Server supplies an AI assistant with reference information for building Paddle integrations. It is aimed at developers looking up current interfaces, reviewing an implementation, or investigating a technical error. It should not be confused with the Paddle MCP Server that performs account operations. A documentation lookup does not require giving an agent permission to change real products or subscriptions. This separation is especially useful when a team is still evaluating an integration. The entry reflects a review of official sources, not an authenticated hands-on test of the hosted service.

Documentation in the agent's working context

According to Paddle, the service covers public developer documentation, the OpenAPI specification, and SDK references for Node.js, Python, Go, and PHP. This gives a development question a closer connection to the published interface than relying on a model's general memory alone. A result still needs to be checked against its original source. When a project uses an older library version, current documentation may not describe exactly the API that is installed. State the SDK language and version before asking for an implementation recommendation.

Where documentation lookup helps

Paddle's official quickstart combines catalog setup, a pricing page, checkout, and webhook handling into an integration workflow. Each step brings narrower questions about parameters, response fields, or event handling. A documentation connector can support that investigation without inspecting the team's production records. A useful review process records the relevant source beside the proposed change and tests the implementation in the project. Treat generated code as a starting point, not as evidence that the complete payment flow has been exercised or that every failure case is handled.

Remote hosting and a separate sign-in

The documented endpoint is https://paddlehq.mcp.kapa.ai, and Paddle identifies Kapa.ai as the hosting platform. Add it as an HTTP MCP connection in the chosen client. The official Claude Code example is claude mcp add --transport http paddle-docs https://paddlehq.mcp.kapa.ai. Initial access requires signing in with a Google or GitHub account, rather than supplying a Paddle API credential. No Paddle account is required. The documentation explains this authentication as a way to enforce rate limits and reduce abuse, and notes that a connection may need to be authenticated again later.

Keep reference access separate from account changes

The Docs MCP Server supplies knowledge, not authorization to edit the team's Paddle business data. A product or subscription operation requires the separate account server with the relevant Paddle authentication and permissions. When both are connected, give them clearly distinguishable names. Ask for an explanation of the documented steps before making a separate decision about executing an account action. This makes it easier to review what was learned, what was proposed, and what a person actually approved.

Privacy and cost considerations

The lookup runs through a hosted service, so it is not a local-only documentation search. Keep confidential customer information, credentials, and internal billing records out of reference questions. A sanitized example describing the operation and a redacted error is usually a better starting point. Returned material becomes available to the AI client, whose processing arrangements must be assessed separately. The reviewed documentation does not list a dedicated Docs MCP usage charge. That is distinct from any charges for the AI client or model and from Paddle's commercial payment-platform terms when a project later processes real transactions.

Who benefits from the Paddle Docs MCP Server?

It fits developers, technical support teams, and people planning an integration before setting up account access. Specify the framework, SDK language, version, and desired behavior, and request the matching official reference. Documentation access reduces the need to manually move reference text between tools, but it does not remove responsibility for verification. Responses may still make wrong assumptions or suggest code that is unsuitable for the application. Check the source and run appropriate project tests. Finally, the license of an open SDK is not evidence that the hosted MCP implementation shares that license; no public implementation repository or open-source license is established in the reviewed service documentation.

A remote HTTP-capable MCP client that can complete Kapa.ai sign-in with Google or GitHub. No Paddle account or Paddle API key is required.

In Claude Code: claude mcp add --transport http paddle-docs https://paddlehq.mcp.kapa.ai. Complete Google or GitHub authentication afterwards. For other clients, configure an HTTP connection to this endpoint and follow the current client-specific sign-in instructions in Paddle's documentation.

claude mcp add --transport http paddle-docs https://paddlehq.mcp.kapa.ai

Kapa.ai sign-in with Google or GitHub for rate limiting and abuse prevention. No Paddle API key. Reauthentication may be required.

Lookup of public Paddle developer resources, OpenAPI specifications, and SDK references. This docs service does not authorize changes to Paddle account data.

Hosted on Kapa.ai. Documentation queries are sent to a remote service and results reach the AI client. The reviewed sources establish neither local-only processing nor specific retention periods. Keep secrets and customer records out of queries.

Confidential information in documentation queries, expired sign-in, rate limits, and unsuitable or misinterpreted reference answers. Check the SDK version and original source; do not confuse the docs service with the write-capable Paddle account connector.