Wander Agent

pip install wander-agent

Try these the second you install — paste any of them to your AI:

• 🛫 "Find me a flight from NYC to Tokyo the airlines don't want me to see." → hidden-city + split-ticket fares OTAs are banned from showing you
• 🌴 "Plan a 10-day Bali trip for 2 from Dubai in September, budget $3,000." → flights, hotels, visa, weather, restaurants, packing list, day-by-day — one response
• 🛂 "Which of my two passports gets me into more countries visa-free?" → dual-passport diff with Henley rankings
• 💳 "Turn my Amex points into the best business-class flight to Europe." → points sweet-spot finder across transfer partners

What is this?

Wander Agent is a tool plugin for AI assistants (Claude, Cursor, Windsurf, etc.) that gives the AI real travel data to work with.

Without this plugin, asking an AI "cheapest flights from New York to Tokyo in August?" gets you a made-up answer. With this plugin, the AI calls real data sources — Google Flights, Open-Meteo, US State Dept advisories, OpenStreetMap — and gives you actual prices, real forecasts, and live visa requirements.

You talk to the AI the same way you always do. The AI does the rest.

"Plan me a 10-day trip to Bali for 2 people leaving from Dubai in September, budget $3,000" → Real flights, real hotels, visa requirements, weather forecast, local restaurants, packing list, day-by-day itinerary — all in one response.

What can it do?

Setup (5 minutes)

Step 1 — Install Python

You need Python 3.10 or newer. Check if you have it:

python3 --version

Don't have it? Download from python.org.

Step 2 — Install Wander Agent

From PyPI (recommended):

pip install wander-agent

Or with uv — no install at all, run directly:

uvx wander-agent

Or from source:

git clone https://github.com/VirajMishra1/wander-agent.gitcd wander-agentpip install -e .

Verify it worked:

wander-agent --help

Step 3 — Connect to your AI

Pick the app you use:

{  "mcpServers": {    "wander-agent": {      "command": "wander-agent"    }  }}

which wander-agent

{  "mcpServers": {    "wander-agent": {      "command": "/usr/local/bin/wander-agent"    }  }}

claude mcp add wander-agent wander-agent

claude mcp list

{  "mcpServers": {    "wander-agent": {      "command": "wander-agent"    }  }}

{  "mcpServers": {    "wander-agent": {      "command": "wander-agent"    }  }}

{  "mcpServers": {    "wander-agent": {      "command": "wander-agent"    }  }}

{  "mcpServers": [    {      "name": "wander-agent",      "command": "wander-agent"    }  ]}

{  "context_servers": {    "wander-agent": {      "command": {        "path": "wander-agent",        "args": []      }    }  }}

Optional: API keys for better data

Everything works without any API keys. These add richer data if you want:

Set them before starting your AI:

export FOURSQUARE_API_KEY=your_key_hereexport TICKETMASTER_API_KEY=your_key_here

Or add them to your shell profile (~/.zshrc, ~/.bashrc) to make them permanent.

All 75 Tools

✈️ Flights

🏨 Hotels

📋 Visas & Entry

🌤️ Weather

⚠️ Safety & News

💰 Costs & Currency

🗺️ Planning & Scoring

🍜 Food & Nightlife

🌐 Smart Trip Tools

🎒 On the Ground

🏛️ Attractions & Info

🧠 Traveler Profile

🧳 Saved Trips (cross-session memory)

Persisted to ~/.wander_agent/trips.json. A trip carries an 8-item booking checklist (flights, hotel, visa, insurance, ground transport, advisory, packing, notify bank).

🔔 Fare Watching

Persisted to ~/.wander_agent/fare_watches.json. Records a baseline price and a price history per watched route.

⚖️ Value Ranking

💳 Credit Card Points & Miles

Static dataset covering 20 loyalty programs (Chase UR, Amex MR, Citi TYP, Capital One, Bilt, 11 airlines, 4 hotel chains), 9 popular credit cards with bonus categories, and 10 curated sweet-spot awards. Card portfolio saved to ~/.wander_agent/profile.json.

Tool Rankings: Usefulness × X-Factor

Ranked by a combination of practical impact and uniqueness — things no OTA, no chatbot, and no other travel tool does.

Data sources

How much can you trust each number?

Most results carry a data_meta block so the AI can tell you how reliable a figure is before you book:

"data_meta": {  "confidence": "scraped_live",  "trust_score": 80,  "trust_label": "Live scrape",  "meaning": "Scraped from Google Flights just now. Real but can shift minute to minute.",  "fetched_at": "2026-06-02T18:30:00+00:00",  "source": "google_flights + kiwi.com"}

Honest limitations

Flight prices: Google Flights scraper can break if Google changes their page structure. Kiwi prices are real and bookable. Both run in parallel — if one fails the other still returns results.

Split-ticket risk: find_split_ticket surfaces real savings but carries missed-connection risk. If your first flight is delayed the second airline won't wait and won't rebook you. The tool always returns a full risk disclosure list. Buy travel insurance that covers missed connections on separate tickets.

Hotel prices: Every hotel price API is paid. This tool returns hotel names and links to 9 booking sites where you can see real prices. No prices are shown directly.

Visa data: Static snapshot (updated 2024). Policies change — always verify with the official link in the response before you book anything. Do not rely on this for immigration decisions.

Transit visa data: Rules encoded as a static dataset covering 200+ combinations. Edge cases (nationalities with bilateral exemptions, recent policy changes) may be missing. Always verify with the airline or destination embassy.

Passport power data: Henley Index 2024 snapshot. Rankings shift year to year. Confirm with official sources before making nationality-based decisions.

SIM card data: Curated snapshot as of 2024-Q4. Prices and plans change frequently. Treat as guidance; verify before purchase.

Events: Scraped from Eventbrite. Best coverage for events 0–60 days ahead. Further-future events may not be listed yet.

Travel advisories: US State Dept only. Updated within 24–48 hours of official changes via RSS.

Health data: Curated snapshot from CDC Yellow Book and WHO 2024–2025 recommendations. Not a substitute for advice from a travel medicine clinic.

Ground transport: Rome2Rio data for route discovery. Actual prices and schedules must be confirmed on the booking sites — times and fares change.

Points/miles valuations: Baseline cpp values are editorial averages (TPG, NerdWallet, OMAAT). Actual redemption value depends on specific award availability, transfer bonuses, and dynamic pricing. Sweet spot awards assume saver-level availability. Card bonus categories reflect terms as of 2025-Q1 — issuers change these periodically.

Carbon calculations: ICAO/DEFRA 2024 emission factors with standard RFI multiplier (×1.9). Actual emissions vary by aircraft type, load factor, and routing. Use as an estimate, not an accounting figure.

Cloud deployment (Docker)

The included Dockerfile and railway.toml let you host this server in the cloud so you can connect to it from any device — not just the computer where you installed it.

You do not need this for normal local use. Install and run locally as described in Setup above.

If you want to self-host on Railway:

# Push to GitHub, then connect repo to Railway# Set env vars in Railway dashboard if using optional API keys

The server switches from stdio (local) to HTTP automatically based on the WANDER_TRANSPORT environment variable.

Development

# Install with dev dependenciespip install -e ".[dev]"
# Run testspython -m pytest tests/ -v
# Count registered toolsgrep -c "@mcp.tool()" src/wander_agent/server.py

Adding a tool:

1. Create src/wander_agent/tools/your_tool.py with an async function
2. Import it in src/wander_agent/server.py
3. Wrap with @mcp.tool() and an async wrapper function
4. Restart your AI client to pick up the new tool

CI: every push and PR to main runs the test suite on Python 3.10/3.11/3.12 and asserts the registered tool count.

Releases: publishing a GitHub release triggers the PyPI publish workflow (trusted publishing — no tokens). Bump version in pyproject.toml and server.json first.

License

MIT

mcp-name: io.github.VirajMishra1/wander-agent