Pipedream MCP
Overview
The Pipedream MCP connector enables your Leena AI Agents to integrate with thousands of third-party SaaS applications through Pipedream's Model Context Protocol (MCP) server, giving your agents broad, agent-native reach across the enterprise without having to build and maintain individual connectors for every downstream tool.
Pipedream is an integration platform that provides managed authentication, credential storage, and a library of pre-built actions for 2,500+ applications including Slack, Gmail, Google Sheets, Salesforce, HubSpot, Jira, GitHub, Notion, and more. The Pipedream MCP connector exposes these applications to Leena AI as MCP "tools" that can be dynamically discovered and invoked by your agents at runtime, using the industry-standard Model Context Protocol.
The full list of apps and connectors available through Pipedream can be viewed here: https://pipedream.com/apps
Each Pipedream MCP connection in Leena AI is scoped to a single Pipedream app (for example, slack or google_sheets). To enable automation across multiple downstream apps, configure one Pipedream MCP connection per target app. Credentials, auth flows, and token refresh for each downstream app are managed entirely by Pipedream.
API Details
Leena AI integrates with Pipedream via its hosted Model Context Protocol (MCP) server, using streamable HTTP/SSE as the transport layer. Tool discovery and invocation conform to the MCP specification.
Documentation link: https://pipedream.com/docs/connect/mcp/developers
Setup
The Pipedream MCP connector supports two authentication modes. The CIO/IT team should choose the mode that best aligns with your governance model:
- Leena Service Account: Credentials are managed entirely by Leena AI. No Pipedream workspace, project, or OAuth client setup is required from your team. This is the fastest path to value and is recommended for most deployments.
- Custom Account: Your organisation brings its own Pipedream workspace and OAuth client. Recommended when you require full control over billing, logging, and data residency for downstream API calls, or when you want to isolate Leena AI activity within your own Pipedream project.
Additionally, both modes support two OAuth flow types for downstream app authorisation:
- Admin OAuth: A single workspace-level credential is used for all end users. Recommended for internal automations where actions run under a central service identity.
- User OAuth: Each end user authorises their own downstream account (for example, their own Gmail or Salesforce). Recommended for employee-facing workflows where actions must run under the individual user's permissions.
Prerequisites
Before setting up the Pipedream MCP connector, ensure you have:
- Access to your Leena AI workspace with connector management permissions
- For Custom Account mode only:
- Administrator access to your Pipedream workspace
- Ability to create projects in the Pipedream workspace
- Ability to create OAuth clients in the workspace API settings
No Pipedream-side setup is required for Leena Service Account mode.
Get credentials
This section applies only to the Custom Account authentication mode. Skip this section if you plan to use the Leena Service Account mode.
Here is how to create a Pipedream project and OAuth client:
- Log in to your Pipedream workspace at https://pipedream.com as a workspace admin
- Create a Project:
- Click Projects in the top left of the Pipedream dashboard
- Create a new project (or select an existing one)
- Copy the Project ID from the project URL (starts with
proj_, for exampleproj_ABC123)
- Create an OAuth Client:
- Navigate to your workspace's API settings
- Click New OAuth Client
- Enter a descriptive name (for example, "Leena AI Integration") and click Create
- Copy the Client Secret (it is displayed only once during creation)
- Click Close
- Copy the Client ID from the OAuth clients list
- Identify the App Slug:
- Decide which downstream Pipedream app this connection will target (for example, Slack, Gmail, HubSpot)
- Note its name slug from the app's page on Pipedream (for example,
slack,gmail,hubspot)
Add connection
Here is how to add a connection on Leena AI:
- Log in to your Leena AI workspace
- Navigate to Settings > Integrations
- Search for "Pipedream MCP" and select it from the list to add a new connector
- Start configuring the connector
- Auth Type: Choose the authentication mode:
- Leena Service Account: Credentials managed by Leena AI (no further credential fields required)
- Custom Account: Bring your own Pipedream workspace credentials
- OAuth Type: Choose how end-user authorisation is handled:
- Admin OAuth: A single set of workspace credentials is used across all users
- User OAuth: Each end user authorises their own downstream account
- App Name: The slug of the Pipedream app this connection will target (for example,
slack,google_sheets,hubspot). Search and select from the list of available Pipedream apps. - Environment: The Pipedream project environment:
- Development: For testing and non-production scenarios
- Production: For live production workloads
- Client ID: The OAuth client ID from your Pipedream workspace (required for Custom Account mode only)
- Client Secret: The OAuth client secret from your Pipedream workspace (required for Custom Account mode only)
- Project ID: The Pipedream project ID, for example
proj_ABC123(required for Custom Account mode only) - Webhook URI: The callback URL used by Pipedream to post events (auto-generated by Leena AI)
- Auth Type: Choose the authentication mode:
- Save the configuration
- Click Connect to validate the credentials and establish the connection
- For User OAuth connections, each end user will be prompted to authorise their own downstream account the first time an action runs on their behalf
- Once validated, the connection will be saved and the available tools for the selected app will be dynamically discovered from the Pipedream MCP server
Actions
Unlike static connectors, the Pipedream MCP connector exposes actions dynamically based on the Pipedream app selected during connection setup. Actions are represented as MCP "tools" and are discovered at runtime from Pipedream's hosted MCP server. When the connection is saved, Leena AI lists the available tools for that app and auto-generates the input form for each so your builders can use them in workflows without additional configuration.
For example:
- A connection configured with App Name =
slackexposes Slack tools such as sending messages, creating channels, and updating user status - A connection configured with App Name =
google_sheetsexposes Google Sheets tools such as adding rows, updating cells, and creating worksheets - A connection configured with App Name =
hubspotexposes HubSpot tools such as creating contacts, updating deals, and adding notes
To view the full list of apps and their supported tools, refer to the Pipedream app catalog at https://pipedream.com/apps.
Common Use Cases
The Pipedream MCP connector can be leveraged by Leena AI Orchestrator/Agent to automate a wide range of cross-system workflows. Here are some common use cases:
- IT Service Management: Create, update, and close tickets in Jira, ServiceNow, Zendesk, or Freshservice based on employee queries
- HR Automation: Push onboarding tasks to Asana, Monday, or ClickUp; sync employee data with HCM systems
- Communication Workflows: Send notifications to Slack, Microsoft Teams, or Gmail when specific events occur in other systems
- CRM Updates: Create contacts, deals, or notes in Salesforce or HubSpot from employee-submitted information
- Document Management: Create or update files in Google Drive, Notion, Confluence, or SharePoint
- Data Capture: Append rows to Google Sheets, Airtable, or databases for auditing and reporting
Action Structure
Every dynamically-loaded tool follows a common structure when configured in Leena AI:
Input Parameters
The input parameters for each tool are defined by the selected Pipedream app and are fetched at runtime via the MCP listTools call. Leena AI renders these as a form, with mandatory and optional fields derived from the tool's input schema.
Mandatory
| Name | Description |
|---|---|
| App Account (Auth) | The connected downstream account (for example, the specific Slack workspace or Google account) that the tool runs against |
| Tool-specific fields | The fields marked as required in the tool's input schema (for example, Channel and Message Text for a Slack "Send Message" tool) |
Optional
| Name | Description |
|---|---|
| Tool-specific fields | The fields marked as optional in the tool's input schema (for example, thread ID, attachments, formatting options) |
Response
Upon successful execution, the tool returns the response payload from the downstream Pipedream app, which typically includes:
- Resource ID of the created, updated, or retrieved entity
- URL to the resource (where applicable)
- Execution status and timestamp
- Full response object from the downstream API
Best Practices
- Start with Leena Service Account mode: Unless you have specific governance requirements, use the Leena Service Account mode for the fastest deployment and lowest operational overhead
- One connection per app: Configure a separate Pipedream MCP connection for each downstream app your agents need to interact with
- Choose OAuth mode carefully: Use Admin OAuth for system-level automations and centralised auditing; use User OAuth when the action must respect individual user permissions and data access scopes
- Separate environments: Use the Development environment for testing new workflows, and Production for live deployments to ensure isolation of credentials and logs
- Rotate secrets regularly: For Custom Account mode, rotate your Pipedream OAuth client secret periodically and update the connection in Leena AI to maintain security posture
- Least privilege: When creating downstream OAuth apps (for example, a Google Cloud OAuth client for Gmail actions), grant only the scopes required for your workflows
- Monitor usage: Track tool invocations in Leena AI logs; in Custom Account mode, also review the Pipedream project dashboard for end-to-end observability
Updated 6 days ago
