Pipedream

What it does

Pipedream today is a product called Connect. It gives developers managed authentication for more than 3,000 APIs, over 10,000 prebuilt tools and triggers, a proxy to call any endpoint, SDKs and a hosted MCP server. Your users connect their own accounts, and Pipedream stores and refreshes the credentials.

There is a second product, Conduit, in early access. It is a gateway that lets employees connect their apps to AI agents, with SSO, access policies, per-user permissions and an audit trail.

Why you would need it

Say you are building an agent that posts to Slack, reads Gmail and updates a CRM on behalf of each customer. Every one of those apps needs an OAuth app registered, tokens stored and refreshed, and permissions handled per user. That is weeks of plumbing before the agent does anything useful.

Pipedream takes that layer off your plate. You call a tool, and the right user's credentials are attached for you.

Where it fits

Pipedream sits between your agent or app and the APIs your users already use. Its docs show it working with the Vercel AI SDK, OpenAI, Anthropic Claude and Google Gemini. It replaces the integration layer you would otherwise build yourself.

It no longer fits as a general automation builder. The vendor has announced that its Workflows product and String shut down on 31 March 2027. If you want a workflow tool, look at Zapier, Make or n8n.

What stands out

• Managed auth. Hosted OAuth, token storage and refresh for more than 3,000 APIs.
• A large tool catalogue. The prebuilt actions and triggers sit in a source-available registry on GitHub, so you can read and extend them.
• A proxy for everything else. If no prebuilt tool exists, you send a request through the proxy and the user's credentials are injected.
• A hosted MCP server. One remote endpoint gives an agent tool discovery and per-user authentication.

My take

I have used Zapier since 2014 and I am a Zapier Certified Expert. I looked at Pipedream as the technical alternative, with code steps in Node.js, Python, Golang or Bash next to the no-code blocks. My view then was based on desk research. I have not run it in production.

Since then Pipedream has changed direction. It has moved away from workflows and towards being the integration layer for AI agents. I would not start a new automation project on the old Workflows product. For an agent product with many connected users, Connect is the part worth a serious look.

Verdict

Pick Pipedream when you are a developer building an agent or app that needs per-user access to many third-party APIs. Skip it if you want a visual workflow builder for your own team, or if nobody on your side writes code.