A small Kamchatka travel platform opens 22 tools to AI agents through MCP
Vedar, a local travel platform in Russia's Kamchatka, lets AI assistants check live tour availability and send booking requests, with real people confirming every order.
The short version
- Vedar's MCP server exposes 22 tools in three groups: lookup, safety and environment, and planning and requests.
- Tools read the platform's own business data, so prices and dates match what agencies maintain in the back office.
- Every create_* tool only files a request; agencies or carriers confirm by phone and no payment happens online.
- Contact details are sent only when the parameter consent: true is set; car rental needs a two-step confirmation.
- The server needs no login because every lookup is read-only.
Why it matters
The post shows that a very small business can make itself visible to AI assistants with a single web endpoint, rather than waiting for a big platform. It also offers a practical pattern for staying safe: let the AI prepare a request, and leave money and final decisions to people.
The problem: AI can't see small shops
Ask a chatbot about volcano hikes in Kamchatka and next week's openings, and you will usually get a general overview plus advice to contact a local agency. According to the developers, the reason is that most agencies there are tiny teams. Their information is scattered across Russian-language sites, Telegram channels and phone calls, and a model's training data has no live prices or free seats.
As dev.to reported in a write-up by the team behind Vedar, they run a local platform that collects tours, transfers, car rentals and lodging from regional agencies. Seeing more travelers plan with AI assistants, they opened the platform to AI agents using MCP, the Model Context Protocol, an open standard for connecting AI to outside tools.
How the 22 tools are organized
The server is remote, uses Streamable HTTP and needs no authentication, since all lookups are read-only. The team chose not to have the AI scrape their web pages. Instead the tools read the platform's business data directly, so the AI sees the same prices and dates agencies maintain, not an outdated page.
- Lookup: tours, details, live availability, lodging, transfers, car rental, gear, and place information.
- Safety and environment: daily weather, volcano status (aviation color codes plus earthquake data), and region-wide warnings and route risk notes.
- Planning and requests: day-by-day trip plans that can be edited later, plus tools to submit and track requests.
Design choices that limit risk
The agents never take payment. Every create_* tool files a request that an agency or carrier confirms by phone. The authors judged that having an AI pay for a user is too risky, while having it organize the user's needs for a human is about right.
Personal contact details are sent only when the call explicitly carries consent. Car rental takes two steps: a draft with a confirmation code, then the user's own approval. Prices must come from a specific catalog tour with a link, because the team considers an invented price worse than none.
Two further choices reflect local knowledge. Trip plans leave a spare day in a week, since helicopter and boat trips are often cancelled for weather. And volcano warnings appear in the same conversation as the related tour recommendation.
A sample trip and what comes next
The authors describe a traveler flying from Harbin to Kamchatka on a Saturday for a week of volcano viewing. An agent could check safety status, find tours with open places, build a Saturday-to-Saturday plan and then send a booking request to an agency. They note that a new Harbin route and visa-free entry for Chinese citizens make this timely.
The post is a developer's own account. It gives no results, usage figures or costs, so how well it works in practice is unknown.
What to watch
- Do travelers and AI assistants actually use the Vedar server, and does it lead to real bookings?
- Will other small, long-tail businesses copy this approach of exposing live data through one endpoint?
Try it
If you run a small business, list the three questions customers ask most, such as availability, prices and bookings, and sketch which could be read-only tools.
Sources and links
- Reportingdev.to: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by dev.to · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
