The travel industry is undergoing a significant infrastructure shift as major players integrate the Model Context Protocol (MCP). KAYAK has launched an agent-to-agent connector, while Amex Global Business Travel (GBT) has integrated directly with Claude’s MCP ecosystem. This move signals a transition from traditional app-based booking interfaces to AI agents that can execute complex travel workflows autonomously. For developers and enterprise IT leaders, this represents a fundamental change in how travel data is exposed and consumed.

From UI Navigation to API Invocation

KAYAK’s new connector allows AI agents to bypass the graphical user interface entirely. Instead of users opening a browser or app to scroll through flight results, an AI can call KAYAK’s MCP interface to run searches, compare prices, and filter options programmatically. This shifts KAYAK from being a destination for human users to a backend tool called by AI. The result is a streamlined experience where the user issues a natural language command, and the system returns the optimal booking option without manual intervention.

Amex GBT Integrates Claude for End-to-End Execution

Amex GBT, a leader in corporate travel management, has taken integration a step further by embedding Claude into its operational flow. Previously, employees had to manually search for compliant flights, submit for approval, pay, and later file expense reports. With the new MCP integration, an employee can simply tell Claude, 'Book me a business trip to Shanghai next Friday.' The AI then interacts with Amex GBT’s interface to search for policy-compliant options, submit approvals, process payment, sync calendars, and generate expense reports post-trip. This automation drastically reduces the administrative burden on both employees and finance teams.

Domestic and Niche Players Join the Ecosystem

The trend is not limited to global giants. Hello Chuxing has adopted MCP-style integrations for ride-hailing, allowing AI assistants to book rides by calling the service’s interface directly. Simultaneously, domestic platforms like RollingGo are building robust hotel MCPs. RollingGo’s solution supports over 40 major LLM clients, including Cursor, Claude Code, and Windsurf, providing access to more than 2 million global hotel properties. By offering direct inventory and real-time price confirmation via MCP, these platforms allow developers to integrate booking capabilities without the high costs of traditional API maintenance.

Key Takeaways

  • MCP adoption is moving travel booking from manual UI interaction to AI-driven autonomous execution.
  • Amex GBT’s integration with Claude automates the entire lifecycle from search to expense reporting.
  • Developers can now access hotel inventory and ride-hailing services via standardized MCP interfaces, reducing integration overhead.
  • Corporate travel policies must be structured and machine-readable to enable AI-driven compliance checks and auto-approvals.

The Bottom Line

The shift from 'people operating platforms' to 'AI calling tools' is no longer theoretical. Corporate travel leaders who fail to structure their policies for machine readability will find their legacy systems incompatible with the incoming wave of AI-driven booking assistants.