mcpField
V
Viajante

Viajante

Local
github.com/felipebasurtoMaps & Travel39Updated 1w ago

About Viajante

Viajante is a local travel-search tool for flights and hotels, aimed at terminal users, Python developers, and AI assistants. It gathers offers from travel sites and reports returned prices and itinerary details without filling in unknown information.

Key features

  1. 01Search flights across dates and routes
  2. 02Search hotels and room rates
  3. 03Compare combined flight and hotel trips
  4. 04Recheck offers and validate itineraries
  5. 05Record and monitor prices optionally

Use cases

  1. 01Compare fares across a date window
  2. 02Find hotel stays with cancellation terms
  3. 03Check a proposed itinerary against search results
  4. 04Track changes in saved search prices

MCP support

Its MCP server provides 22 tools for flight, hotel, itinerary, airport, and price searches. It can run over stdio or Streamable HTTP; the HTTP transport has no authentication and should stay on loopback.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"felipebasurto-viajante": {
"command": "npx",
"args": [
"-y",
"@viajante/mcp"
]
}
}
}
Run commandnpx -y @viajante/mcp

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
pypiviajante1.5.2
npm@viajante/mcp1.5.2

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

Viajante: flight and hotel search for your terminal, Python, and AI assistants. Runs on your machine, no API keys.

Flight and hotel search for your terminal, Python, and AI assistants. Runs on your machine, no API keys.

PyPI Python npm Tests

Website · Usage guide · Architecture · Changelog

Four points: runs on your machine with no API keys; returns evidence, not advice; only provider_empty means nothing was found; three ways in, a terminal command, a Python call, or an AI assistant over MCP.

It reads Google Flights, Google Hotels, and Booking.com directly, and Skiplagged when you opt in. It returns prices, itinerary details, and links. It reports what the provider returned and says so when something is unknown; it never invents a fare.

Quick start

Requires Python 3.10+.

pip install viajante

viajante flights JFK-LHR:2026-11-15 --fetch sweep
viajante hotels London 2026-11-15 2026-11-20 --currency GBP --source google

Use future dates. Not sure of a code? viajante airports london.

Prefer not to install? uvx --from viajante viajante airports JFK, or npx -y -p @viajante/mcp viajante airports JFK (both need uv).

These searches use plain HTTP. Booking.com and --fetch detail need the browser extra.

Connect an AI assistant

Add this to your assistant's MCP configuration (needs uv and Python 3.10+):

{
  "mcpServers": {
    "viajante": {
      "command": "npx",
      "args": ["-y", "@viajante/mcp"]
    }
  }
}

Without Node, use "command": "uvx" with "args": ["--from", "viajante[mcp]", "viajante-mcp"]. That covers Google Flights and Hotels with no Chromium. The npm entry also includes the browser extra, which search_explore needs.

Then ask in plain language:

Find a seven-night round trip from BOS to LHR, departing any day in November 2026. Compare the dates.

Hotels in Tokyo, November 12 to 16, two adults, free cancellation, in JPY.

Every tool result (except lookup_airports) carries one envelope: status, completeness, empty_reason, retry_after, observed_at. Only empty_reason: provider_empty means the provider found nothing. A client that only takes a URL can use viajante-mcp --transport streamable-http; it has no auth, so keep it on loopback (details).

All 22 MCP tools
ToolWhat it does
search_flightsOne-way, round-trip, or multi-city flights
search_datesCheapest fare per departure date in a window
search_flexDates around a target, then flights for the cheapest day
search_exploreDestinations from an origin, with a priced shortlist
search_hotelsStays with total-stay prices and cancellation terms
search_hotel_roomsSkiplagged room rates for one named hotel
get_hotel_detailsA hotel offer from this session, plus an optional room quote
search_tripFlights and hotel together, summed when compatible
search_split_ticketsOpt-in separately ticketed itineraries (connections not protected)
search_hidden_cityOpt-in Skiplagged hidden-city fares
recheck_offerOne fresh search to re-check an earlier offer
validate_itineraryCheck a proposed itinerary against search evidence
compare_awardsCents-per-point math for a named award offer
lookup_transfersPoints transfer-partner table
lookup_airportsAirport and metro codes, offline
plan_stay_blocksGroup consecutive nights with the same people
split_stay_costsSplit stay totals among who slept there
verify_answerFlag amounts, codes, dates, or links no search returned
price_historyPrices this machine recorded for a query (opt-in)
watch_priceRe-run a saved search and report the change
get_runtime_infoInstalled version, offline
get_guideLong operational guide

Command line

Run viajante <command> --help for options. Add --save FILE to write a JSON report.

CommandPurpose
flightsSpecific routes and dates (--nearby-days compares ±N days when N is set; --split-tickets is opt-in)
datesCheapest departure day across a window of up to 31 days
flexA few days either side of a target date
exploreDestinations from an origin airport
hotels, hotel-roomsGoogle Hotels, Booking.com, or Skiplagged
tripFlights and a hotel stay in one request
recheck-offerRe-check a saved offer against a fresh search
airportsFind airport codes
hidden-city, awards, pointsSkiplagged fares, award value, transfer partners
history, watchRecorded prices and saved re-runs (opt-in)
feedbackOpen a prefilled GitHub issue (viajante sends nothing itself)
viajante dates BOS-LHR --from 2026-11-01 --to 2026-11-30 --nights 7

Python

from viajante import get_flights

report = get_flights("JFK-LHR:2026-11-15", fetch="sweep", top=5)

for result in report.queries:
    if result.status == "ok":
        for offer in result.offers:
            print(offer.price, report.currency, offer.airline, offer.duration)
    else:
        print(result.error)

FlightQuery, HotelQuery, and the search_* functions are exported too; see the Python examples.

Optional browser support

pip install 'viajante[browser]'
playwright install chromium

For uvx, use viajante[mcp,browser] in --from and install Chromium from that same environment.

Good to know

  • Currency. Taken from --currency, or from a named origin airport's country. Hotels always need one. Viajante never converts between currencies.
  • Baggage. --carry-on asks Google to price one carry-on. Checked bags (--bags) are refused, because no transport can verify them. --baggage-buffer only affects ranking; it is not a quoted fee.
  • Hotels. Prices are for the whole stay. Free cancellation is required unless you opt out. Google rates 0–5, Booking.com 0–10.
  • Recommendation. Flight results may include a short recommendation block. It adds to the offer list and never replaces it.
  • Price level. Flight results may include Google's own price level (low, typical, or high) from the results page already fetched. It is Google's label for that price against its typical range. --price-history adds the page's price history. No extra request.
  • Unknown stays unknown. Missing fields are left out or marked unknown.

Travel sites change their pages, block requests, and update prices at any time. Confirm the final fare, baggage, hotel total, and cancellation terms with the provider before you book.

Contributing

Bug reports and focused pull requests are welcome; see CONTRIBUTING.md.

License

MIT. Independent project, not affiliated with Google, Booking.com, or any airline. Use of those services is subject to their terms: Google, Booking.com.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.