Fli (Google Flights MCP) — Use Cases, Install & Live Demo

Why use it

Key features

Uses Google Flights' own protobuf APIs — fast and stable, not HTML scraping
Round-trip, one-way, multi-city all supported
Date-grid search: 'cheapest day in May from SFO to Tokyo'
Filters by airline, alliance, stops, cabin, max price
No API key, no signup — works out of the box
Available as Python library, CLI, or MCP server

Live Demo

What it looks like in practice

flights-fli-mcp.replay ▶ ready

0/0

Install

Pick your client

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "flights-fli-mcp": {
      "command": "uvx",
      "args": [
        "fli-mcp"
      ]
    }
  }
}

Open Claude Desktop → Settings → Developer → Edit Config. Restart after saving.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "flights-fli-mcp": {
      "command": "uvx",
      "args": [
        "fli-mcp"
      ]
    }
  }
}

Cursor uses the same mcpServers schema as Claude Desktop. Project config wins over global.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "flights-fli-mcp": {
      "command": "uvx",
      "args": [
        "fli-mcp"
      ]
    }
  }
}

Click the MCP Servers icon in the Cline sidebar, then "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "flights-fli-mcp": {
      "command": "uvx",
      "args": [
        "fli-mcp"
      ]
    }
  }
}

Same shape as Claude Desktop. Restart Windsurf to pick up changes.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "flights-fli-mcp",
      "command": "uvx",
      "args": [
        "fli-mcp"
      ]
    }
  ]
}

Continue uses an array of server objects rather than a map.

~/.config/zed/settings.json

{
  "context_servers": {
    "flights-fli-mcp": {
      "command": {
        "path": "uvx",
        "args": [
          "fli-mcp"
        ]
      }
    }
  }
}

Add to context_servers. Zed hot-reloads on save.

claude mcp add flights-fli-mcp -- uvx fli-mcp

One-liner. Verify with claude mcp list. Remove with claude mcp remove.

Use Cases

Real-world ways to use Fli (Google Flights MCP)

How to find the cheapest day to fly between two cities with Claude

👤 Travelers with flexible dates ⏱ ~10 min beginner

When to use: You know where and roughly when, but you'd rather move the date than pay $200 more.

Prerequisites

fli-mcp installed — uvx fli-mcp (auto-installs from PyPI)

Flow

Ask for a price grid

Use Fli. Show me the cheapest day to fly SFO → NRT in the second half of June, round-trip, 7-day stay.✓ Copied

→ Date grid table with daily round-trip lows
Drill into the cheapest pair

For the cheapest date pair, list the actual itineraries — airlines, layovers, total time.✓ Copied

→ 5-10 itineraries sorted by price, with airline + duration
Narrow down

Filter to non-stop only, max 1 layover, depart after 9am from SFO.✓ Copied

→ Re-filtered list, much shorter

Outcome: A short, real-price flight list you can book directly on the airline site or Google Flights.

Pitfalls

Booking site prices differ from what you saw — Google Flights prices are upstream cache; for OTA fees, click through to the airline. Fli matches Google Flights, not the OTA.

Plan a multi-city business trip with Claude doing the searches

👤 Frequent business travelers, EAs ⏱ ~20 min intermediate

When to use: You have 3+ cities to hit and rolling your own search across each leg is tedious.

Flow

Lay out the route

Plan multi-city: SFO → JFK May 12, JFK → LHR May 15, LHR → SFO May 20. One adult, business class, max 1 stop per leg.✓ Copied

→ Multi-city itinerary with priced legs
Compare alliances

Re-run with Star Alliance only so I can keep my UA status accruing.✓ Copied

→ Filtered set, possibly higher price — Claude flags the delta
Export

Save the chosen itinerary to a markdown brief with booking URLs and seat policies.✓ Copied

→ Brief written to disk; URLs work

Outcome: A booked-ready multi-city plan in under 10 minutes instead of 45 min of tab-juggling.

Pitfalls

Some airlines are missing from Google Flights (Southwest) — Fli inherits Google Flights' coverage. Southwest must be checked separately.

Combine with: filesystem

Build a flight-deal watchlist Claude refreshes on demand

👤 Hobbyists and digital nomads ⏱ ~25 min intermediate

When to use: You want to track 5–10 routes and get told when a price dips.

Flow

Define the watchlist

Track these 5 routes for the next 60 days, store each search as JSON in /flights/watch.json.✓ Copied

→ watch.json created with 5 routes
Run the sweep

Fetch fresh lowest prices for every route in watch.json. Diff against last run, highlight drops > 15%.✓ Copied

→ Diff table; drops bolded

Outcome: A lightweight personal price-drop alert without paying Hopper or Going.

Pitfalls

Hammering Google Flights too fast gets temporary blocks — Space the sweep — 1 query per ~3 sec is fine. Run it on a cron not a tight loop.

Combine with: filesystem

Combinations

Pair with other MCPs for X10 leverage

flights-fli-mcp + filesystem

Persist a flight watchlist and price history to /flights/

Run today's watchlist. Save results to /flights/$(date +%Y%m%d).json and append to history.csv.✓ Copied

flights-fli-mcp + google-calendar

Once an itinerary is chosen, drop the legs onto your calendar

Add this round-trip's outbound and return as Calendar events with airport codes in the title.✓ Copied

Tools

What this MCP exposes

Tool	Inputs	When to call	Cost
search_flights	origin: str, destination: str, depart_date, return_date?, passengers, cabin	Standard one-way or round-trip search	1 Google Flights query
search_date_grid	origin, destination, month, trip_length	When dates are flexible and you want lowest-day discovery	1 query (heavier)
search_multi_city	legs[]: {origin, destination, date}	3+ city itineraries	1 query
filter	itineraries, filters: {airlines?, alliance?, max_stops?, time_window?, max_price?}	Narrow a result set without a re-query	free (post-fetch)

Cost & Limits

What this costs to run

API quota: No formal quota — Google Flights enforces soft per-IP limits (~1 query per few seconds)
Tokens per call: 300–2500 tokens per result set
Monetary: Free
Tip: Use search_date_grid sparingly — it's heavier than point queries.

Security

Permissions, secrets, blast radius

Credential storage: None — no API keys

Data egress: All requests go to flights.google.com

Google could change the protobuf schema at any time. The library is updated when that happens but not under SLA.

Troubleshooting

Common errors and fixes

HTTP 429 / temporary block

You queried too fast. Wait 5–10 minutes and slow your sweep cadence.

Verify: Re-run a single query after the wait

Empty result for valid route

Date may be > 11 months out (Google's max). Pick a date inside that window.

Some airline missing (Southwest etc.)

Google Flights does not index Southwest. Use airline.com directly for those.

Alternatives

Fli (Google Flights MCP) vs others

Alternative	When to use it instead	Tradeoff
SkyScanner / Amadeus / Duffel APIs	You're building a real product and need TOS-compliant booking	Those need keys + payment. Fli is best for personal automation, not a public app.
Kiwi.com Tequila API	You need explicit aggregator coverage including budget carriers	Free tier limited; Fli has zero quota but Google Flights coverage.

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills