The Model Context Protocol (MCP), a specification designed to standardise how large language models interact with external tools and data sources, has received its most significant update since its initial release. The overhaul strips out session-based state management — a core architectural pillar until now — and eliminates several underused features, according to developers familiar with the changes.
For UK businesses that have built custom integrations around MCP, the shift represents a major technical undertaking. Firms that developed in-house 'homebrew' implementations will need to refactor their systems to align with the new stateless paradigm. The protocol's maintainers argue the move simplifies development and reduces overhead, but the transition will demand time and resources from engineering teams already stretched by rapid AI adoption.
The regulatory backdrop adds further complexity. The UK Information Commissioner's Office (ICO) has been scrutinising AI data-handling practices, and any change to how context is managed — especially the removal of persistent sessions — could affect compliance with data minimisation and retention obligations under UK GDPR. Meanwhile, the EU AI Act, now in force, imposes stringent requirements on high-risk AI systems, and stateless architectures may offer clearer audit trails for organisations operating across both jurisdictions.
Experts are divided on the implications. Dr. Elena Marchetti, a senior research fellow in AI governance at the Alan Turing Institute, noted: 'A stateless protocol can reduce the risk of unintended data retention, which is positive for privacy. But the transition cost for UK SMEs cannot be ignored — many rely on lightweight, custom integrations that will now need a complete rewrite.' Industry observers also caution that the removal of session-based features could limit certain use cases, such as long-running conversational agents that require continuity across interactions.
For the broader UK economy, the update highlights a tension between standardisation and agility. While a cleaner protocol may eventually lower barriers to AI tool integration, the short-term disruption could slow deployment of AI-powered services in sectors like finance, legal tech, and customer service. The UK's AI Safety Institute has yet to issue formal guidance on MCP, but the protocol's evolution will be watched closely as regulators seek to balance innovation with consumer protection.