Trvl Mcp: One Tool, 65 Intents, Zero API Keys

One Tool to Route Them All

Most travel MCP servers drown you in endpoints. Trvl Mcp takes the opposite approach — one smart tool, 65 ways to use it, and zero API keys standing between you and real travel data.

That's not a limitation. That's a design philosophy.

The entire server is built around a single tool: travel. You pass it an intent — a string that can be one of 65 aliases — and it routes your request to the right data source automatically.

Flights? Hotels? Trains? Ferries? Lounges? It's all handled by the same interface. Under the hood, trvl-mcp queries Google Flights, Google Hotels, Trivago, Airbnb, Booking.com, Hostelworld, Ferryhopper, and European ground transport networks — all without you ever touching an API key.

The intent aliases cover an impressively wide surface area:

• Air travel: flights, price alerts, award sweet spots, baggage rules, lounges
• Accommodation: hotels, Airbnb, hostels via Hostelworld, Booking.com properties
• Ground transport: rental cars, trains, buses, ferries across European networks
• Trip intelligence: destination intel, weather, and more

The award_sweet_spots intent alone is worth the install for frequent flyers. Finding where your miles stretch furthest is notoriously painful — having an AI agent that can surface that data on demand, mid-conversation, is genuinely useful.

At a total score of 88/100, trvl-mcp punches above its weight for a server with only 33 GitHub stars. Let's look at where those points come from.

The security score is a full 30 — the ceiling. That's not an accident. By abstracting away credentials entirely, the project sidesteps a whole class of problems that plague other data-fetching MCP servers. You don't need to manage tokens, rotate keys, or worry about leaking credentials through your AI client's context window.

There's a real architectural opinion embedded in this server. Most MCP tools give you granular, separate tools for each action — one for flights, one for hotels, one for weather. That's orthodox MCP design.

trvl-mcp bets that intent routing is actually better for AI agents. Instead of the model deciding which tool to call, it just describes what it wants and the server figures out the rest. 65 aliases act as compatibility shims for natural language variations — flights, flight, find_flights, search_flights all resolve to the same behavior.

That bet has tradeoffs. You get less granular control per-intent, and debugging a routing failure is harder when there's only one tool to inspect. But for the use case — an AI travel agent that just works — the simplicity wins.

Travelers building personal AI assistants will get the most mileage here. If you're running Claude or another LLM locally and want it to actually help with trip planning — not just hallucinate flight options — this gives it real data to work with.

Developers prototyping travel features benefit enormously from the zero-credential setup. You can test the full stack — flights, hotels, ground transport — without a single API account.

Frequent flyers and points optimizers should pay special attention to the award_sweet_spots intent. It's a niche feature that signals the server was built by someone who actually travels.

The 33 GitHub stars suggest this is still early. But the 88/100 score reflects that the fundamentals are solid — clean architecture, zero credential friction, and a broad data source network that most travel tools can't match without significant setup.

If you've been waiting for an MCP travel agent that doesn't require you to become a developer before you can search for flights — this is the one to install.