Skip to content

MCP Server

The Profinity Model Context Protocol (MCP) server lets an AI tool read live data from one Profinity instance. The server answers questions about components, tag values, tag history, and alerts, and has no tool that changes Profinity data. The AI tool sees only what its account is allowed to see.

AI Chat uses this server inside Profinity. Claude Code, Cursor, and Codex connect to it from your computer. With AI Skills installed, those tools can then build dashboards, rules, and other configuration from what they read.

Licence required

The MCP server requires the AI licensed feature, included in the Server and Enterprise editions. Refer to Licensing for what each edition includes.

Choose a setup guide

Prepare the instance once, then connect the tool you use:

  1. Prepare Profinity for an MCP client: turn the server on, create a token, choose an address, and check that the server answers.
  2. Trust the HTTPS certificate: needed only when the tool uses an https:// address.
  3. The guide for your tool: Claude Code, Codex, or Cursor.

ChatGPT cannot reach a private MCP server, so use a Custom GPT and the REST API instead. The AI Skills reference gives the reasons. To use ChatGPT with Profinity, refer to Connect ChatGPT to Profinity.

What the server returns

The server's tools cover three jobs:

  • Find components and tags.
  • Read alert rules, active alerts, and alert history.
  • Read current values and history.

Every call needs a token for an account with the MCP integration permission. Tag tools also need View tags, and alert tools also need View alerts. The answer silently includes only the tags and alerts that the account can view.

The MCP tool reference lists each tool and the permission it needs. When a connection fails or returns less than you expect, refer to MCP connection errors.