Model Context Protocol (MCP)
An open protocol that lets an AI assistant discover and call the tools and data a system exposes — increasingly how audit systems make themselves usable by agents.
MCP standardises the connection between a language model and an external system. Instead of every vendor writing a bespoke integration for every assistant, a system publishes a set of tools — query this data, upload this document, propose this change — and any MCP-speaking client can use them.
For audit the significance is not the protocol itself but what it forces into the open. To expose an audit system over MCP you must be able to say precisely what an agent may read, what it may change, and what requires approval. Systems built around documents struggle to answer that, because their unit of work is a file rather than a fact.
It also changes where the work can happen. An auditor can sit in whichever assistant they already use, reach into the audit system through MCP, and have proposed changes flow back through the normal approval path. The audit platform stops being a place you visit and becomes a service your tools call — which is a considerable demotion for the incumbents whose value rests on being the place you visit.
Related