Skip to main content

Prerequisites

Before installing the MCP server, you’ll need:
  1. Pica Account - Create a free account
  2. Pica API Key - Get your API key from Settings > API Keys
  3. Connected Integrations - Connect integrations from the Connections page
  4. MCP-Compatible IDE - Claude Desktop, Cursor, Windsurf, or any MCP client

Installation

The easiest way to install the Pica MCP Server is via NPM:
npm install -g @picahq/mcp
Alternatively, you can use it directly with npx without installation:
npx @picahq/mcp

Other Installation Methods

Remote MCP Server

Install via remote MCP server

Configuration

Set Your API Key

The MCP server requires your Pica API key to authenticate requests. Set it as an environment variable:
export PICA_SECRET=your_pica_secret_key
Get your API key from the Pica dashboard.

Optional: Identity Scoping

You can scope connections to a specific identity (e.g., a user, team, or organization) by setting these optional environment variables:
export PICA_IDENTITY=user_123
export PICA_IDENTITY_TYPE=user
VariableDescriptionValues
PICA_IDENTITYThe identifier for the entity (e.g., user ID, team ID)Any string
PICA_IDENTITY_TYPEThe type of identityuser, team, organization, project
When set, the MCP server will only return connections associated with the specified identity. This is useful for multi-tenant applications where you want to scope integrations to specific users or entities.

IDE Setup

To use the MCP server in Cursor:
  1. Open Cursor and go to Cursor Settings
  2. Select MCP Settings from the menu
  3. Add the following configuration to your MCP JSON file:
{
  "mcpServers": {
    "pica": {
      "command": "npx",
      "args": ["@picahq/mcp"],
      "env": {
        "PICA_SECRET": "your-pica-secret-key",
        "PICA_IDENTITY": "user_123",
        "PICA_IDENTITY_TYPE": "user"
      }
    }
  }
}
PICA_IDENTITY and PICA_IDENTITY_TYPE are optional. Only include them if you want to scope connections to a specific identity.
  1. Save the file and restart Cursor
Cursor MCP Settings
Testing in Cursor:Try asking Cursor:
  • “What connections do I have access to?”
  • “Show me all Gmail actions I can perform”
  • “Send an email using Gmail to hello@picaos.com
  • “Build a React form component that sends emails with Gmail”

Examples

Once set up, you can use the MCP server for three main purposes:

Direct Action Execution

Ask your AI to perform actions on integrated platforms:
Get my last 5 emails from Gmail
The AI will use the MCP server to fetch your recent emails and display them.
Send a message to #general channel in Slack: "Meeting in 10 minutes"
The AI will post the message to your Slack workspace.
Create a calendar event for tomorrow at 2pm titled "Product Review"
The AI will create the event in your Google Calendar.
Create a lead in Salesforce for John Doe at Acme Corp with email john@acme.com
The AI will create the lead in your Salesforce CRM.

Advanced Configuration

Manual Installation

If you want more control over the installation:
# Clone the repository
git clone https://github.com/picahq/mcp.git
cd mcp

# Install dependencies
npm install

# Build the server
npm run build

# For development with auto-rebuild
npm run watch
Manual IDE Configuration:
{
  "mcpServers": {
    "pica": {
      "command": "node",
      "args": [
        "/path/to/pica-mcp-server/build/index.js"
      ],
      "env": {
        "PICA_SECRET": "YOUR_PICA_SECRET_KEY",
        "PICA_IDENTITY": "user_123",
        "PICA_IDENTITY_TYPE": "user"
      }
    }
  }
}
PICA_IDENTITY and PICA_IDENTITY_TYPE are optional. Only include them if you want to scope connections to a specific identity.

Deploy to Vercel

You can deploy the MCP server to Vercel for remote access:
  1. Install dependencies including Vercel adapter:
    npm install @vercel/mcp-adapter zod
    
  2. Deploy to Vercel:
    vercel
    
  3. Set environment variables in Vercel:
    • PICA_SECRET (required)
    • PICA_IDENTITY (optional)
    • PICA_IDENTITY_TYPE (optional)
  4. Configure your MCP client to use the remote server:
    • For Cursor: https://your-project.vercel.app/api/mcp
    • For Claude/Cline: Use npx mcp-remote https://your-project.vercel.app/api/mcp
See the DEPLOYMENT.md file in the repository for detailed instructions.

Troubleshooting

Problem: The AI can’t connect to the MCP server.Solutions:
  • Verify your PICA_SECRET is set correctly in the configuration
  • Check that the MCP server is running (restart your IDE)
  • Ensure you have an active internet connection
  • Try running npx @picahq/mcp manually to test
Problem: Getting 401 or authentication errors.Solutions:
Problem: The server reports no connections available.Solutions:
Problem: Actions fail to execute.Solutions:
  • Check the action knowledge first using get_pica_action_knowledge
  • Verify all required parameters are provided
  • Ensure your connection has proper permissions for the action
  • Check for rate limits or API-specific restrictions

Next Steps

Claude Desktop Setup

Specific setup guide for Claude Desktop

Browse Integrations

Explore all 200+ available integrations

API Reference

Learn about the underlying Pica APIs

GitHub Repository

View source code and contribute

Get Help

Contact Support

Have questions? Email us at support@picaos.com for assistance