Whop MCP

The API MCP is the server for live actions. Whop documents two remote transports: HTTP Streaming athttps://mcp.whop.com/mcpand SSE athttps://mcp.whop.com/sse. Use this when the agent needs to list products, inspect payments, create checkout configurations, review memberships, or call another Whop API endpoint.

The npm package is the official local Whop API MCP server. It supports normal explicit tools,--tools=dynamicfor large API surfaces, and--tools=codefor a docs search tool plus Deno-sandboxed code execution against the TypeScript client. It also supports--client=cursor,--client=claude,--client=claude-code, and--client=openai-agentscompatibility tuning.

export WHOP_API_KEY="your_api_key_here" export WHOP_WEBHOOK_SECRET="optional_webhook_secret" export WHOP_APP_ID="optional_app_id" npx -y @whop/mcp@latest --client=cursor --tools=dynamic

Use Docs MCP for implementation help and API MCP when the agent needs to test live API calls. Keep the two servers named separately in your client.

Use API MCP only. Start with listing and auditing tasks before approving anything that changes products, plans, payments, memberships, or checkout data.

Use local@whop/mcp --tools=dynamicso the client can discover endpoint schemas instead of loading every tool at once.

Use local@whop/mcp --tools=codewhen the agent needs to write small scripts against the Whop SDK and chain multiple API calls.

Use separate server names for docs and API access. That makes approval dialogs easier to read and reduces the chance that a documentation question turns into a live API call.

Open Claude Web, go to Settings, choose Connectors, add a custom connector, and use the Whop API MCP URL:

When the connector asks for authorization, provide the Whop API key that matches the data and permissions you want the agent to access.

Editclaude_desktop_config.json. On macOS, check~/Library/Application Support/Claude/claude_desktop_config.json. On Windows, check%APPDATA%\Claude\claude_desktop_config.json. Whop's remote setup usesmcp-remoteto bridge the SSE endpoint into Claude Desktop.

Apps, app builds, access tokens, account links, webhooks, files, companies, authorized users, users, and app infrastructure.

Products, plans, experiences, checkout configurations, promo codes, invoices, setup intents, and payment methods.

Payments, refunds, disputes, dispute evidence, dispute alerts, resolution center cases, payment retries, voiding, and fees.

Memberships, members, entries, cancellations, pauses, resumes, access checks, leads, reviews, and shipments.

Courses, chapters, lessons, course students, course lesson interactions, assessments, progress, and content operations.

Chat channels, messages, reactions, support channels, forums, forum posts, direct channels, notifications, and moderation workflows.

Affiliates, affiliate overrides, ad campaigns, ad groups, ads, conversions, promo codes, and growth reporting.

Ledger accounts, payout accounts, payout methods, transfers, withdrawals, topups, fee markups, verifications, and company token transactions.

Stats describe, metric queries, raw queries, and SQL-style analysis when exposed by the active server/tool mode.

Use sandbox first when you are testing checkout, payment, or membership flows. Whop documents a sandbox dashboard and sandbox API base URL athttps://sandbox-api.whop.com/api/v1. The remotemcp.whop.compage does not currently advertise a separate sandbox MCP URL, so the conservative path is to run@whop/mcplocally with a sandbox API key.

export WHOP_API_KEY="your_sandbox_api_key" npx -y @whop/mcp@latest --tools=dynamic

A write-capable MCP server with a real Whop API key can change products, create checkout configurations, refund payments, cancel memberships, send messages, or perform other business actions depending on scopes. Treat it like any production integration.
- Use least-privilege keys.Grant only the scopes needed for the workflow.
- Start in sandbox.Test checkout and payment behavior away from live customers.
- Separate docs from API access.Use Docs MCP for docs questions and API MCP only when live actions are needed.
- Read approval dialogs.Inspect the tool name and arguments, not just the agent's summary.
- Filter tools.Use--tools=dynamic,--operation=read,--resource=..., or exclusion flags to reduce the available tool surface.
- Do not commit secrets.Use environment variables, prompted inputs, or client secret storage.
- Pin and review dependencies.Whop documents@latestandmcp-remote@latestfor convenience, but production workflows should review package changes before upgrading.

Check JSON syntax, full-quit the client, and confirm Node.js 18 or newer is available to the application. GUI apps on macOS sometimes launch with a different PATH than your terminal.

Whop's remote page recommends clearing the MCP auth cache and restarting the client. Remove the~/.mcp-authfolder with your operating system's file manager or another safe deletion method, then reopen the MCP client.

A 401 usually means the token is missing or invalid. A 403 usually means the key does not have the permission required by the endpoint. Re-check the key type and the endpoint's documented scopes.

Use a more specific prompt, split docs and API servers, and filter unnecessary tools. Include the target resource, date range, action type, and whether writes are allowed.

Use@whop/mcp --tools=dynamicor filter with resource/operation flags. Large APIs can overwhelm clients when every endpoint is exposed as a separate tool.

The official Whop server is the right default when you want a direct Whop API or documentation connection. Hosted gateways make sense when you want scheduling, no-code orchestration, multi-app workflows, or a client that does not support MCP directly.

The Whop MCP server connects MCP-compatible AI clients to Whop documentation or Whop API tools. The Docs MCP helps agents read documentation. The API MCP can call live Whop API endpoints with your API key.

There are two official server targets in Whop's docs:https://docs.whop.com/mcpfor documentation andhttps://mcp.whop.com/mcporhttps://mcp.whop.com/ssefor API actions.

It is the official remote Whop API MCP server landing page. It documents setup for Claude Web, Claude Desktop, Cursor, and Windsurf using the remote SSE endpoint.

It is the official npm package for running the Whop API MCP locally. It supports explicit tools, dynamic endpoint discovery, code mode, client compatibility flags, filtering, and HTTP transport.

Use a Company API key for your own company data, an App API key for an installed Whop app, and OAuth tokens when acting on behalf of an individual user. Use the smallest permission set that works.

Yes, when the active API MCP exposes checkout configuration tools and your key has the required checkout, plan, and access-pass permissions. Ask the agent to show arguments before approving a write call.

Does Whop MCP support affiliate tracking?

Whop checkout APIs and embeds support affiliate attribution through fields such asaffiliate_codeandaffiliateCode. Whop's affiliate guide also documents referral links with?a=<username>.

How do I add Whop MCP to Claude Desktop?

Add anmcp-remoteserver entry pointing tohttps://mcp.whop.com/sseinclaude_desktop_config.json, then fully restart Claude Desktop.

Addhttps://docs.whop.com/mcpfor docs lookup, or usemcp-remote@latestwithhttps://mcp.whop.com/ssefor API access. Use global or project-levelmcp.json.

Is it safe to give MCP write access to my Whop store?

It can be safe with the right controls, but it is not casual access. Use sandbox first, limit key permissions, filter tools, and review every write-capable tool call before approving it.

Use a direct MCP client where possible. For clients without arbitrary MCP support, route through a hosted gateway and review how that gateway stores credentials and executes actions.

Use these as the source of truth when details change: