Pica

You can run and interact with the Pica MCP Server to connect to many platforms, execute actions directly, and generate integration code within a secure, scalable MCP interface. This server centralizes platform connections, action execution, and code generation through a consistent API exposed to MCP clients and tooling.

How to use

You will use an MCP client to discover available integrations, locate actions for specific platforms, view detailed action knowledge, and execute actions through the MCP interface. Start by listing your integrations, then search for platform actions, review their requirements, and finally execute actions when you are ready. You can also leverage automatic code generation for common workflows and UI components that interact with platforms.

Typical usage patterns include: selecting a platform to view its actions, inspecting an action’s parameters and usage, and performing the action with the necessary data. All requests pass through a secure proxy, and secrets are never exposed in responses. You can scope connections to a specific identity to support multi-tenant setups.

How to install

Prerequisites: ensure you have Node.js and npm installed on your machine.

Install the MCP server package from npm.

npm install @picahq/mcp

Additional configuration and usage notes

Environment variables you will commonly set when running the server include a secret for Pica and optional identity scoping keys. You can run the server locally in standard environments or connect remotely to a hosted MCP server.

Optional identity scoping lets you restrict connections to a specific user, team, or organization. This is useful for multi-tenant applications where you want to isolate integrations per identity.

Security and authentication are handled through Pica's secure proxy. No direct platform API keys are exposed, and sensitive data is filtered from logs.

If you plan to use Claude Desktop or Cursor, you will configure the MCP server in their respective config files. Examples below show how to reference the MCP server from a local runtime and how to set identity context when needed.

Remote access is also available through a hosted MCP endpoint. Use the provided URL to connect your MCP client to the remote server.

Examples and configurations

Claude Desktop and Cursor integrations use a local MCP server configuration to run the MCP client as a background service or within the host editor. The following examples illustrate how to point clients at the MCP server and how to scope identity if required.

{
  "mcpServers": {
    "pica": {
      "command": "npx",
      "args": ["@picahq/mcp"],
      "env": {
        "PICA_SECRET": "your-pica-secret-key",
        "PICA_IDENTITY": "user_123",
        "PICA_IDENTITY_TYPE": "user"
      }
    }
  }
}

To run the MCP server directly, you can use the standalone CLI invocation shown here.

npx @picahq/mcp

Available tools

list_pica_integrations

List all available Pica integrations and platforms. Use this first to discover platforms and connections.

search_pica_platform_actions

Search for relevant actions on a specific platform using a query. Returns up to 5 top actions.

get_pica_action_knowledge

Get comprehensive documentation for a specific action, including parameters and usage details.

execute_pica_action

Execute a Pica action to perform operations on third-party platforms. Returns a sanitized request config and actual API response.