The new Higson MCP Server is a built-in bridge between your AI assistant and your live rules engine.

MCP (Model Context Protocol) is the emerging open standard for connecting AI assistants to the systems where real work happens: letting tools like Claude Code, etc securely read, understand, and act within your systems instead of guessing from the outside.

What it means in real life?:

• Explore and search every element in your configuration
• Create and edit decision tables, Groovy/Python functions, domain models, context types and flows
• Test before you publish run test cases and read the results
• Approve before it goes live - changes are staged for explicit sign-off

Because business rules are too important to hand off blindly, control is built in: AI changes are staged, then reviewed and published by a person - never applied on their own. Access runs through a secure token you manage.

There are dedicated skills available for the MCP Server that make it easier to work with Higson and speed up the assistant’s implementation in everyday tasks. To learn how to enable and configure the MCP Server, see MCP Server in the Technical Documentation.

The higson-plugins repository bundles a Higson skill with ready-made MCP server configuration for Claude Code and OpenAI Codex, letting your coding agent read, change, test, and publish Higson decision tables, functions, domain, and flows through natural-language requests.