Phocas MCP server
Use Model Context Protocol (MCP) to connect AI assistants to Phocas favorites data.
This page is for developers and other technical readers who already know MCP. It skips the basics and focuses on how to use MCP with Phocas.
The experimental Phocas MCP server currently supports two tools:
ListFavorites
Returns a list of all favorites available to you.
GetFavoriteData
Returns the data for a specified favorite. Standard user restrictions are applied to the dataset.
Get started
Administrators must assign the API tokens > MCP (Model Context Protocol) permission to appropriate user profiles.
Create a token
API Tokens are created per user.
The MCP server respects your settings and permissions, and only shows favourites you can access. If restrictions apply, the returned data is filtered accordingly.
For detailed steps, see Manage API tokens.
Configure your agent
Use these settings when you connect your agent to the Phocas MCP server:
Authentication: Bearer Token
Test the MCP server in your agent
We don't replicate third-party setup instructions here, because those products change often. You might also have several agents to choose from in the same product.
Use the vendor documentation for the full steps. The notes below will help you get started.
Claude Code CLI
Use this command to add the Phocas MCP server to Claude Code CLI:
claude mcp add --transport http phocas-mcp https://api.phocassoftware.com/mcp --header "Authorization: Bearer [your_token]"
ChatGPT
Option 1: In platform.openai.com
Open Chat.
Next to Tools, click Add.
Select MCP Server, then Server.
Enter the URL and token, then click Connect.
Option 2: OpenAI Agent Builder
Add the Phocas MCP server as part of OpenAI Agent Builder tool.
Gemini CLI
For setup steps, see https://geminicli.com/docs/tools/mcp-server/
Understand the limitations
The GetFavoriteData payload is limited to 6 MB, and there is currently no pagination.
Make sure your favorite returns less than this amount of data. If it does not, you will receive an error.
Last updated