Product Changelog

​

MCP Chat - Talk directly with all your tools

• It’s powered by Pipedream Connect and Vercel’s AI SDK and works with a wide variety of LLMs
• Direct access to >10k tools from >2800 APIs (with auth built-in of course)
• We’ve open sourced MCP Chat (MIT license), so you can fork it or use it as a reference implementation for your own AI app

Summarize my recent emails

Draft a product release announcement based on my team’s recent Linear tickets

Find the latest screenshot in my Google Drive and send it to my team on Slack

Help me prep for my next meeting — who am I meeting with and what are we discussing?

​

New Developer Playground for Pipedream Connect

• We added a Debug tab so you can inspect all of the calls happening in realtime (with payloads and responses)
• Fixed bugs and expanded connect-react capabilities (added support for sql and object prop types and improved the UX for string[] inputs)
• We gave it a big overall facelift with a fresh new design throughout the playground
• The code is public — check out the code for the playground and run it locally here

​

Remote MCP Server and OpenAI Support

​

Remote MCP server for developers

• You no longer need to host and deploy Pipedream’s MCP server in order to add it to your app (you still can if you prefer though)
• You can now make authenticated requests to https://remote.mcp.pipedream.net
• Check out our docs for more info

​

MCP support in OpenAI’s Responses API

• OpenAI just released native support for using MCP servers in their Responses API
• We’re excited to have partnered with them to make sure Pipedream has a drop-in solution to use our MCP server with their SDK
• Check out the docs to get started

​

Code examples

// Import the Pipedream SDK
import { createBackendClient } from "@pipedream/sdk/server";
 
// Initialize the Pipedream SDK client
const pd = createBackendClient({
  environment: PIPEDREAM_ENVIRONMENT,
  credentials: {
    clientId: PIPEDREAM_CLIENT_ID,
    clientSecret: PIPEDREAM_CLIENT_SECRET,
  },
  projectId: PIPEDREAM_PROJECT_ID
});
  
// Get access token for MCP server auth
const accessToken = await pd.rawAccessToken();

import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";

const serverUrl = MCP_SERVER_URL || `https://remote.mcp.pipedream.net/${externalUserId}/notion`;
 
const transport = new StreamableHTTPClientTransport(new URL(serverUrl), {
  requestInit: {
    headers: {
      "Authorization": `Bearer ${accessToken}`,
      "x-pd-project-id": PIPEDREAM_PROJECT_ID, // proj_xxxxxxx
      "x-pd-environment": PIPEDREAM_ENVIRONMENT // development | production
    }
  }
});

import OpenAI from 'openai';
 
// Initialize OpenAI client
const client = new OpenAI();
 
// Make the OpenAI request with the MCP server
const response = await client.responses.create({
  model: 'gpt-4.1',
  tools: [
    {
      type: 'mcp',
      server_label: 'Notion',
      server_url: `https://remote.mcp.pipedream.net/${externalUserId}/notion`,
      headers: {
        Authorization: `Bearer ${accessToken}`,
        "x-pd-project-id": PIPEDREAM_PROJECT_ID,
        "x-pd-environment": PIPEDREAM_ENVIRONMENT
      },
      require_approval: 'never'
    }
  ],
  input: 'Summarize my most recently created Notion doc for me and help draft an email to our customers'
});
 
console.log(response);

​

Pipedream Connect Updates

​

MCP

• ICYMI, we published all 2600+ integrated apps as MCP servers for individuals to use for free and for developers to deploy to their AI apps and agents
• We published an update to the MCP server reference in the public repo
• We released support for the new streamable HTTP transport type (in addition to SSE and stdio)
• We added a debug flag to make it easier for devs to see what API calls are happening from tool calls (pd_sdk_debug=true)
• We added support for configuring dynamic props
• Check out the code

​

Pre-built tools

• query is now a supported option in configureComponent, so your end users can easily search across remote options for Google Spreadsheets, Notion docs, and more
• Email, HTTP, and Schedule triggers are now supported in Connect (docs)
• Basic source logs are now available for deployed triggers that are owned by external users (docs)
• We published an API for updating deployed triggers (docs)
• Check out the docs

​

Connect Proxy

• The Connect proxy now supports nearly 100% of all integrated apps and we significantly improved the developer interface to remove a lot of the complexity: as a dev, you don’t need to reference auth at all — the proxy will automatically inject the necessary headers or params where they need to go, based on the upstream API’s requirements
• We increased the timeout window from 2 to 30 seconds
• Check out the docs

​

Overall Connect API, SDK, and Docs

• Added an interactive managed auth quickstart to help you get started more easily
• Improved type definitions
• Added featured_weight to the /apps API (docs)
• Shipped various bug fixes
• And we released a much lower priced plan to use Connect in production

​

Connect your AI assistant to any API

• Claude, Cursor, Windsurf, and many AI SDKs all have built-in support for MCP servers in their desktop apps, and OpenAI just announced they’re rolling out support as well
• Chatting with an LLM no longer has to be siloed from the rest of your work — add MCP servers for Slack, GitHub, Linear, Gmail, etc, and go wild!
• Enable your AI assistant to access your Notion workspace so it can get up to speed on your internal processes and knowledge base, then use it to actually draft and reply to customer requests via Front or Intercom

• MCP is a new standard that Anthropic co-authored and is promoting within the AI industry, which provides a common interface for LLMs to communicate with APIs and other resources
• The MCP spec is nascent but quickly evolving, and there’s a lot of excitement from many developers within the AI community
• Generally, the idea is that it makes it easier to provide LLMs and AI agents with specific tools to accomplish tasks

​

Send custom API requests for your users

1. Avoid dealing with authorization grants or token storage and refresh with managed auth to let your customers connect to >2500 APIs
2. Get started more quickly by accessing hundreds of approved Pipedream OAuth clients in production
3. Send custom requests to any integrated API and Pipedream automatically inserts the required auth header or token for your end user

​

Add 10k tools to your AI agent with Connect

• Connect to every app: Get immediate access to 10,000+ tools across 2,500+ apps so you can run any action and deploy any trigger directly from your AI agent.
• No more auth headaches: Pipedream manages OAuth token storage and refresh so you don’t have to worry about managing user credentials — you can even use our pre-approved OAuth clients to avoid lengthy approval processes.
• Full UI Control: Choose between our pre-built UI components (via the connect-react package) or build your own using our server-side SDK.

​

What’s possible with the Connect SDK and API?

• AI agents: Seamlessly power your agent with the tools they need to get their tasks done. With over 2,500+ apps supported, we have you covered.
• Embedded workflow builders: Creating an AI agent solution built around a workflow experience? Use the Connect SDK to power your integrated tools. You control the UI, we handle the integrations.
• Product integrations: Not focused on AI agents but still looking to rapidly expand integration coverage for your SaaS app? The Connect SDK and API enables new integrations to be added in a few hours instead of days/weeks.

​

What’s coming next

​

Run workflows for your end users

​

Getting started

1. Use the Connect frontend SDK or Connect Link to let your users connect their account
2. Within any workflow, toggle Use end user’s auth via Connect on any step to use your end user’s auth instead of your own
3. Invoke your workflow via HTTP trigger or use Pipedream’s SDK
4. Learn more in our quickstart

​

What’s next

​

Introducing Pipedream Connect

​

Managed authentication

• Connect to all 2,400+ supported APIs, directly from your application
• Retrieve end user credentials via API
• Free on all plans up to 1,000 connected accounts
• Check out the docs

• Use the Pipedream SDK or Connect Link to handle authorization or accept API keys for any of Pipedream’s 2,400+ supported APIs.
• Securely retrieve OAuth access tokens, API keys, and other credentials for your end users with Pipedream’s REST API.

1. Open an existing project or create a new one
2. Click on the Connect tab on the left
3. Click on the Users tab at the top
4. Click Connect account

​

Coming next

• Run Pipedream workflows on behalf of your end users
• Embed any action or trigger directly in your application
• Join us on Slack to learn more and request early access

​

Parallel branching now available in beta!

​

More intuitive and powerful filtering

​

New and improved test options

• Test individual steps
• Test entire workflow, segment or block
• Test from the start of the workflow or segment
• Test to the end of the workflow or segment

​

Switch operator for conditional branching now in beta!