Pipedream MCP Server

The Pipedream MCP Server gives your AI assistant access to more than 3,000 apps and APIs, from Slack and Gmail to Notion and Linear, Google Sheets, HubSpot or GitHub. MCP, the Model Context Protocol, is an open standard that lets assistants such as Claude, ChatGPT or Cursor use external services. The server comes from Pipedream, an integration platform that, by its own account, is used by more than a million developers. In November 2025, Workday announced it would acquire the company. Pipedream runs the server itself in the cloud. You install nothing; you simply add an address to your client.

What is the Pipedream MCP Server good for?

Most official MCP servers cover exactly one product. If your day involves ten different services, you would have to set up ten servers, maintain ten sign-ins and keep ten entries in your configuration. Pipedream brings all of that into one place. According to the company, more than 10,000 ready-made tools are available, such as sending a Slack message, adding a row to Google Sheets, creating a Linear ticket or drafting an email in Gmail. The tools come from the same library of actions that Pipedream has used for its automations for years. That makes the server especially useful for apps that do not have an MCP server of their own yet.

Two routes: personal use and your own agents

Pipedream offers the server in two flavors. For personal use, you add the address mcp.pipedream.net/v2 to your client. The first time you connect, you sign in to Pipedream MCP, pick the apps you want and grant access. Note that this account is separate from a regular pipedream.com account, according to the documentation. The second flavor is for developers who want to build MCP into their own product or AI agent. It runs through a separate endpoint on Pipedream Connect. Your app fetches an access token and, with every request, states which project, environment, end user and app the tools are meant for. If a user has not connected an app yet, the server returns a link to connect it right in the response.

Managed sign-in to every app

Authentication is a major advantage. According to Pipedream, it handles the full OAuth flow for every app. OAuth is a sign-in method where you allow access in your browser without handing over a password. Pipedream stores the credentials encrypted and isolated per user. The documentation states that they are never exposed to the AI model or to client-side code. Connections can be revoked at any time. For developers, this removes the most tedious part of any integration: registering your own OAuth apps with dozens of services and keeping tokens refreshed.

Supported clients

Pipedream documents dedicated setup steps for Claude as a custom connector, for ChatGPT in developer mode, and for Cursor and VS Code with one-click installation. Any other MCP-capable client, such as Claude Code, simply uses the server address in its configuration, according to the docs. One practical tip from the documentation: the server name you choose becomes part of every tool name. Some clients cap the length of those names, so a short name like PD helps.

Free to start, paid plans for production

According to Pipedream, the hosted MCP servers are free for personal use. Developers can also build and test for free in development mode. A paid plan only becomes necessary once your own app goes to production with Pipedream Connect. The Startup plan adds a production environment with an allowance of end users and credits. The Business plan adds SSO, SCIM, HIPAA support and dedicated static egress IPs. Current terms are listed on the Pipedream pricing page.

Source code and license

The MCP server itself is a hosted service and has not been published as a standalone open source project. Its tools, however, are generated from the public component library in the PipedreamHQ/pipedream repository on GitHub. That repository is covered by the Pipedream Source Available License. So the code is open to read, but it is not an open source license in the usual sense.

Who is the Pipedream MCP Server for?

The server makes sense for anyone who wants to connect an assistant to many services at once without maintaining a separate server for each. That includes knowledge workers who handle email, calendars, spreadsheets and project tools from a single chat. It is even more interesting for teams building their own AI agent or SaaS product who want to offer integrations to their customers. Compared with Make or Zapier, the focus is not on triggering prebuilt workflows. Instead, the AI calls individual actions directly. That is more flexible, but it calls for more care with permissions. If you only need one or two services that already have a good official server, using those directly is often the simpler path.

For personal use, a free Pipedream MCP account and a client that supports remote MCP, such as Claude, ChatGPT, Cursor, VS Code or Claude Code. Claude and ChatGPT need a plan that supports custom connectors. For your own apps, a Pipedream project with a client ID and client secret; production needs a paid Connect plan.

With the Pipedream MCP Server, your AI assistant can reach more than 3,000 apps without you setting up a separate server for each one. Pipedream hosts the server, so you only add an address. This guide covers setup in common clients and the developer route.

Account and apps

The first time you connect, you sign in to Pipedream MCP or create an account. This account is separate from a regular pipedream.com account. You then pick the apps the assistant may use and grant access to each. On mcp.pipedream.com you can browse the available apps and tools in advance.

Claude

Open the connector settings in Claude and choose Add Custom Connector. Enter the address https://mcp.pipedream.net/v2 and click Add. Then allow access and enable Pipedream in the chat settings. Custom connectors require a Claude plan that supports them.

ChatGPT

In the connector settings, turn on developer mode under Advanced Settings. Click Create, enter the server address and allow access. In the chat, pick Pipedream via the plus sign, More and Developer Mode.

Cursor and VS Code

For Cursor version 1 or later and VS Code with MCP support, Pipedream offers one-click installation. After that, you only need to allow access. Without that button, add an entry called pipedream with the url https://mcp.pipedream.net/v2 to your MCP configuration.

Claude Code and other clients

Every MCP-capable client uses the same address. In Claude Code, the command claude mcp add --transport http pipedream https://mcp.pipedream.net/v2 is all you need. On first use, your browser opens for sign-in. Choose a short server name, because it becomes part of every tool name and some clients cap their length.

For developers: your own agents

If you want to build MCP into your own app, use Pipedream Connect and the endpoint https://remote.mcp.pipedream.net/v3. Your app fetches an access token with its client ID and client secret and sends it as a bearer header. Add the headers x-pd-project-id, x-pd-environment, x-pd-external-user-id and x-pd-app-slug. If a user has not connected an app yet, the response contains a link to connect it. For testing, the MCP Inspector works well via npx @modelcontextprotocol/inspector.

Work safely

Connect only the apps you actually need. Many tools write or send data, so have your client confirm those actions. You can revoke unused connections at any time in your Pipedream MCP account.

claude mcp add --transport http pipedream https://mcp.pipedream.net/v2

Personal use: sign in to Pipedream MCP in the browser, then OAuth per app, managed by Pipedream. Developers: client credentials OAuth via api.pipedream.com/v1/oauth/token, with the access token sent as a bearer header and the end user identified by an external user ID.

Depending on the connected app, read and write whatever the respective action allows, such as sending messages, adding spreadsheet rows, creating tickets or drafting emails. The scope follows from the permissions you grant when connecting each app.

According to Pipedream, credentials are stored encrypted and isolated per user and are not passed to the AI model. App content flows through the Pipedream cloud and, in the chat, also reaches the respective AI provider. Connections can be revoked at any time.

Medium to high. A single service holds access to many apps, and many tools write or send data. A crafted text with hidden instructions can trigger unwanted actions. Connect only the apps you really need, have your client confirm write actions and revoke connections you no longer use.