lastminute.com, a European travel-tech company, said it launched its first Model Context Protocol (MCP) Server for Flights, describing it as the first step in a broader infrastructure roadmap that also includes planned MCP servers for hotels and dynamic packages.
The company said the flight MCP server is already live and listed on the Anthropic Marketplace, where developers can begin experimentation.
What the server does
In its announcement, lastminute.com said the server provides structured, API-level access that lets large language models “such as Claude” query its flight inventory. The company described the release as building “AI-ready” access to core travel services, rather than adding a conversational layer on top of existing systems.
CEO Alessandro Petazzi said the launch is a “direct and tangible result” of lastminute.com’s AI roadmap and pointed to “agentic” travel as a shift in how trips will be planned and booked.
What MCP is
MCP was introduced by Anthropic in November 2024 as an open standard for connecting AI applications to external data sources and tools via MCP servers and clients. OpenAI’s documentation describes MCP as a standard way to connect models to tools and context and notes support for MCP servers in Codex tooling.
Agentic adoption signals
The shift to agentic systems among enterprise players is on the rise. Gartner predicted in 2025 that 40% of enterprise applications will include integrated task-specific AI agents by 2026, up from less than 5% in 2025.
Deloitte reported in its 2025 Emerging Technology Trends study that 38% of surveyed organizations are piloting agentic solutions, while 11% said they are actively using them in production, and it also pointed to the need for “FinOps” controls to manage agent-driven costs tied to token-based pricing.
PwC’s 2026 AI predictions similarly argued that companies are moving toward agentic AI benchmarks and tighter measurement, saying spend is shifting toward initiatives with proof points and metrics that track business value, while governance is struggling to keep pace with adoption.
Travel technology providers have also been positioning MCP as an “agentic-ready” interface layer. Sabre has described “agentic-ready” APIs enabled by its own MCP server, positioning it as a way for AI agents to discover tools and orchestrate travel tasks.
Expedia Group’s developer documentation says it is “building MCP capabilities” and positions MCP servers as a way to streamline travel-business operations through AI agents.
Security and governance
Security and governance are part of the deployment calculus. Thoughtworks describes MCP-Scan as a security scanner for MCP servers designed to detect issues such as prompt injection and “toxic flows,” including through proxy-mode monitoring.
Microsoft has also published guidance on indirect prompt injection risks in MCP-enabled setups and mitigations for MCP scenarios.