mcpField
N
navifare-mcp

navifare-mcp

Local
github.com/navifareMaps & Travel0Updated 29w ago

About navifare-mcp

navifare-mcp is an MCP server that checks whether a flight a traveler has already found is available for less. It parses flight details and compares prices across booking sources, supporting round trips with the same origin and destination.

Key features

  1. 01Parse flight details from natural language
  2. 02Compare prices across booking sources
  3. 03Check a previously found flight
  4. 04Support round trips with matching airports

Use cases

  1. 01Compare a quoted fare with other sources
  2. 02Turn conversational flight details into a price-check request

MCP support

Exposes tools to parse flight information and check prices to MCP clients. The production endpoint is https://mcp.navifare.com/mcp.

Configuration

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

mcp.json · Universal config
{
"mcpServers": {
"simonenavifare-navifare": {
"command": "npx",
"args": [
"-y",
"navifare-mcp"
],
"env": {
"GEMINI_API_KEY": "<GEMINI_API_KEY>",
"NAVIFARE_API_BASE_URL": "<NAVIFARE_API_BASE_URL>"
}
}
}
}
Run commandnpx -y navifare-mcp
Environment variables to set
GEMINI_API_KEYSecretGoogle Gemini API key for image extraction (required)
NAVIFARE_API_BASE_URLNavifare API base URL (optional, defaults to https://api.navifare.com/api/v1/price-discovery/flights)

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
npmnavifare-mcp0.1.3

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

Navifare MCP Server

Version: 0.1.5

Navifare finds a better price for a specific flight the user already found. Users should provide flight details conversationally, which will be structured into the required format.

Overview

This MCP server wraps the Navifare REST API as MCP tools for flight price comparison. It enables AI assistants like Claude to help users find better prices for flights they've already discovered.

MCP Endpoint

  • Production: https://mcp.navifare.com/mcp
  • Documentation: https://www.navifare.com/mcp (Auto-updates from the MCP endpoint)
  • Learn more about MCP: Model Context Protocol

Tools

format_flight_pricecheck_request

Title: Format Flight Request

Parse and format flight details from natural language text or transcribed image content. Extracts flight information (airlines, flight numbers, dates, airports, prices) and structures it for price comparison. Returns formatted flight data ready for flight_pricecheck, or requests missing information if incomplete.

Annotations:

  • readOnlyHint: true - Tool only formats/parses data, no external calls
  • destructiveHint: false - Tool does not modify or delete data

Input Schema:

  • user_request (required): Flight details in natural language text. Include all available information: flight numbers, airlines, departure/arrival airports and times, dates, prices, passenger counts, and travel class.

Example Request:

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "format_flight_pricecheck_request",
    "arguments": {
      "user_request": "I found an Aeroitalia flight XZ2020 from FCO to MXP leaving tomorrow at 19.31 and returning with XZ2021 MXP to FCO next Friday at 10.11. 1 adult, economy, the best fare was 221 EUR"
    }
  }
}
flight_pricecheck

Title: Flight Price Check

Search multiple booking sources to find better prices for a specific flight the user has already found. Compares prices across different booking platforms to find cheaper alternatives for the exact same flight details.

Annotations:

  • readOnlyHint: false - Tool performs searches and may trigger external API calls
  • destructiveHint: false - Tool does not modify or delete data

Input Schema:

  • trip (required): Flight trip details including segments, passengers, and travel class
    • legs (required): Array of flight legs (one for outbound, one for return in round trips)
    • travelClass (required): ECONOMY, PREMIUM_ECONOMY, BUSINESS, or FIRST
    • adults (required): Number of adult passengers (minimum: 1)
    • children (required): Number of child passengers (minimum: 0)
    • infantsInSeat (required): Number of infants requiring a seat (minimum: 0)
    • infantsOnLap (required): Number of infants on lap (minimum: 0)
  • source (required): Source identifier for the query (e.g., "Claude", "MCP", "ChatGPT")
  • price (required): Reference price found by the user (e.g., "84.00", "200.50")
  • currency (required): Three-letter ISO currency code (e.g., "EUR", "USD", "GBP")
  • location (required): Two-letter ISO country code for user location (e.g., "ES", "IT", "US"). If unsure, default to "ZZ"

Example Request:

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "flight_pricecheck",
    "arguments": {
      "trip": {
        "legs": [
          {
            "segments": [
              {
                "airline": "XZ",
                "flightNumber": "2020",
                "departureAirport": "MXP",
                "arrivalAirport": "FCO",
                "departureDate": "2025-12-16",
                "departureTime": "07:10",
                "arrivalTime": "08:25",
                "plusDays": 0
              }
            ]
          }
        ],
        "travelClass": "ECONOMY",
        "adults": 1,
        "children": 0,
        "infantsInSeat": 0,
        "infantsOnLap": 0
      },
      "source": "MCP",
      "price": "84",
      "currency": "EUR",
      "location": "IT"
    }
  }
}

Current Limitations

  • Round-trip flights only: One-way trips are not yet supported. Requests with a single leg will return an error immediately.
  • Same origin/destination: Open-jaw trips (returning to a different airport) are not yet supported. The return flight must depart from the outbound destination and arrive at the outbound origin.

Revenue Share

Revenue share available for qualified partners!

Earn commission on flight bookings made through your integration. Revenue share available for qualified partners.

Compliance

This server complies with Anthropic's MCP Directory Policy. See docs/COMPLIANCE.md for details.

Privacy Policy and Terms of Service: https://navifare.com/terms
Contact: [email protected]
Privacy Inquiries: [email protected]

Environment Variables

The only environment variable you typically need to document for this MCP server is:

  • GEMINI_API_KEY (required): Google Gemini API key for natural language parsing (used when your integration relies on Gemini for text/image understanding).

Examples

See docs/USAGE_EXAMPLES.md for complete usage examples with detailed workflows.

Documentation

Most documentation is in the docs/ directory (and a few files like this README.md and REPOSITORY_STRUCTURE.md live at the repo root):

Getting Started
Reference
Additional Guides

Testing

Test files are located in the test/ directory. See test/README.md for details about available tests.

Running Tests
# Test with MCP Inspector
npx @modelcontextprotocol/inspector node dist/index.js

# Run specific test
node test/test-mcp.js

Scripts

Helper scripts are in the scripts/ directory:

  • scripts/start-local.sh - Start local development server
  • scripts/deploy-to-render.sh - Prepare for Render deployment
  • scripts/start-servers.sh - Start multiple servers
  • scripts/stop-servers.sh - Stop running servers

Repository Structure

navifare-mcp/
├── src/              # TypeScript source code
├── dist/             # Compiled JavaScript (generated)
├── docs/             # Documentation
├── test/             # Test files
├── scripts/          # Helper scripts
└── [config files]    # package.json, Dockerfile, etc.

See REPOSITORY_STRUCTURE.md for detailed structure.

Support

Reference

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.