Integration glossary
MCP server
An MCP server is a service that offers a system's tools and data to AI assistants through the Model Context Protocol, an open standard, so any compatible assistant can use them within the limits the server sets.
What the Model Context Protocol is
The Model Context Protocol (MCP) is an open standard for connecting AI assistants to the systems they work with. An assistant that speaks MCP, such as Claude, Claude Code, Cursor or VS Code, can discover the tools a server offers, call them and read the answers.
Before MCP, each assistant needed its own plug-in for each system. With a shared protocol, a system publishes one server and every compatible assistant can use it.
What a good MCP server controls
Connecting an AI assistant to business systems raises the same questions as giving a new colleague access.
-
Who is calling, with a key for each person or tool.
-
Which tools and which data each key may use.
-
Whether a change is applied directly or waits for someone to approve it.
-
A record of every call, including the ones that were refused.
The STRAX MCP server
The STRAX MCP server serves two kinds of work. Developers connect their AI coding agent to build workflows, sites and reports through proposals that are checked and tested before they run. Business users connect Claude or another assistant to ask about customers and payments, answered read-only within the limits set for each key. Every call is written to the audit trail.
MCP for developers → MCP data access → The AI copilot and MCP, in the documentation →
More terms
Other ideas worth knowing.
Soft delete and delete propagation
Marking a record deleted while keeping the row, and passing a delete on to the other systems that hold the same record.
Read the definition →Worker pool and horizontal scaling
A group of machines sharing one queue of work, so capacity grows by adding another machine to the pool.
Read the definition →Report theme
A saved set of colours, fonts, logo and cover design that every report using it is rendered with.
Read the definition →See it on your own systems.
A demo takes about an hour and shows these ideas working on systems like yours.