Eight Tabs and a Spreadsheet
It was 11 PM on a Wednesday. She had been planning a ten-day trip to Japan for three hours. One browser tab showed Google Flights. Another showed Booking.com. A third had a Google Sheet where she was manually entering hotel prices in column B and converting yen to TWD in column C. Tab four was a weather forecast for Osaka. Tab five was a blog post titled "Hidden Gems in Kyoto" from 2019 — half the restaurants had closed. Tab six was Google Maps, measuring walking distances between temples. Tab seven was an article about Japan Rail Passes that contradicted tab eight, another article about Japan Rail Passes.
She had not booked anything. She had not decided anything. She had organized herself into paralysis. The spreadsheet had forty rows and no conclusions.
This is the modern trip planning trap. Information is not the problem — there is too much of it. The problem is synthesis. You need a system that can take "I want to go to Japan for ten days in April, I care about food and temples, and my budget is $2,500" and turn that into a concrete, day-by-day plan with real numbers attached.
That system exists. It runs in your terminal.
What You Will Build
A travel planner that works like this: you tell the AI where you are going, when, what you care about, and how much you want to spend. It builds a structured itinerary — day by day, with time blocks, location names, estimated costs, and notes. Then it checks live weather forecasts for your travel dates, adjusts the plan accordingly (rainy day on Day 3? Move the outdoor hike to Day 5), calculates a full budget breakdown, and generates a packing list based on the weather and activities.
The output is a markdown file. You can read it in any text editor, print it, commit it to a git repo, or open it in Obsidian. And here is the part that matters: because it is a file, not a chat message, you can edit it. Change your mind about Day 4? Delete those lines, type what you want instead, and ask the AI to adjust everything downstream — budget, packing list, schedule.
The plan is a living document. Not a screenshot of a conversation you will never find again.
What you need:
- Claude Code installed and authenticated. If you have not set it up yet, the First Hour with Claude Code tutorial takes about ten minutes.
- A Claude Pro subscription ($20/month) or an Anthropic API key.
- Basic terminal comfort — if you can type
cdandls, you are ready.
No programming experience required.
Step 1: Set Up Your Travel Project (2 Minutes)
Open your terminal. Create a folder for your trip and start Claude Code inside it:
mkdir japan-april-2026 && cd japan-april-2026
claude
That is it. You now have an AI agent running in a folder dedicated to your trip. Everything it creates will live here — itinerary, budget, packing list, notes.
Step 2: Describe Your Trip in Plain Language (5 Minutes)
Talk to the AI the way you would talk to a travel-savvy friend. Be specific about what matters to you. Vague inputs produce vague plans. Specific inputs produce plans you can actually follow.
Here is an example prompt:
Plan a 10-day trip to Japan starting April 12, 2026.
Travelers: 2 adults.
Budget: $2,500 USD total (excluding flights).
Interests: food (especially ramen, sushi, street food), temples, walking neighborhoods, one day hike.
Base cities: Tokyo (4 nights), Kyoto (3 nights), Osaka (2 nights).
Constraints: no car rental. Public transit only. One rest day built in.
Create a file called itinerary.md with:
- Day-by-day schedule with morning/afternoon/evening blocks
- Specific restaurant and site recommendations with estimated costs in USD
- Daily cost estimate
- Running budget total
- Notes on transit between cities (which trains, rough cost, travel time)
Press Enter. The AI will create itinerary.md in your project folder. Open it in any text editor or markdown viewer.
What you get back is not a list of "top 10 things to do in Tokyo." It is a structured schedule. Day 1, morning: arrive at Narita, take the Skyliner to Ueno (cost: $25), check in at hotel. Afternoon: walk Yanaka neighborhood, visit Nezu Shrine. Evening: ramen at Fuunji in Shinjuku (est. $12/person). Daily total: $120. Running total: $120 of $2,500.
The AI draws on its training data — restaurant recommendations, temple opening hours, train routes, typical costs. It is not perfect. Some prices will be estimates. Some restaurants may have changed. But the structure is solid, and you now have something concrete to verify and refine, instead of eight browser tabs and a blank spreadsheet.
Step 3: Add Live Weather Data (5 Minutes)
An itinerary without weather is a guess. You scheduled an outdoor market for Day 6, but Day 6 might be a typhoon. The AI can check real weather forecasts and adjust.
In the same Claude Code session, type:
Check the weather forecast for Tokyo (April 12-15), Kyoto (April 16-18), and Osaka (April 19-21). Use the Open-Meteo API — it's free, no API key needed.
Then update itinerary.md:
- Add a weather summary line to each day (high/low temp, rain probability)
- If any day has >60% rain chance, swap outdoor activities with indoor alternatives
- Update the packing list section based on the actual forecast
The AI will write a script that hits the Open-Meteo API — a free, no-authentication weather API — fetches the forecast for your exact dates and cities, and weaves the data into your itinerary.
Now Day 3 does not just say "Visit Fushimi Inari." It says "Visit Fushimi Inari — 18C, 15% rain chance" or "Rain expected (72%) — swap to Nishiki Market and Manga Museum (indoor alternatives)."
This is the difference between a plan and a real plan. Weather data turns wishes into decisions.
Step 4: Build the Budget Breakdown (5 Minutes)
You already have daily cost estimates in the itinerary. Now make them rigorous:
Create a file called budget.md with:
- A table: category (accommodation, food, transit, activities, misc) x day
- Row totals and column totals
- A buffer row (10% of total) for unexpected costs
- Currency note: show both USD and JPY at current exchange rate
- Flag any day where spending exceeds $300
Also update itinerary.md to reference budget.md for the detailed breakdown.
The AI creates a clean markdown table. You can see at a glance that Day 2 is your most expensive day ($310 — that sushi omakase) and Day 7 is your cheapest ($85 — the rest day). The buffer gives you $250 of breathing room.
This budget is not locked. It is a markdown file. When you change the itinerary, you ask the AI to recalculate the budget. The numbers follow the plan, not the other way around.
Step 5: Generate the Packing List (3 Minutes)
A packing list should not be generic. It should reflect what you are actually doing and what the weather will actually be.
Create packing-list.md based on:
- The weather forecast already in itinerary.md
- The activities planned (hiking day, temple visits, city walking)
- The trip duration (10 days)
- Two travelers
Organize by category: clothing, electronics, documents, toiletries, misc.
Add notes where relevant (e.g., "comfortable walking shoes — you're averaging 18,000 steps/day based on the itinerary").
The AI looks at the itinerary, sees that you have a hiking day (Day 5 — trail shoes needed), multiple temple visits (no sleeveless tops), ten days of city walking (blister prevention), and April weather in Kansai (layers, light rain jacket). The packing list reflects your specific trip, not a generic "what to pack for Japan" blog post.
Step 6: Iterate — The Living Document (Ongoing)
This is where the terminal approach destroys the browser tab approach.
Your friend messages you: "Skip Osaka, spend those days in Hiroshima instead." In the browser-tab world, you start over. New hotel search, new restaurant research, new train times, new budget calculations. An hour of work, minimum.
In the terminal:
I'm changing the plan. Replace Osaka (Days 8-9) with Hiroshima.
Update itinerary.md, budget.md, and packing-list.md.
Keep the same daily budget target.
Add the Hiroshima Peace Memorial and Miyajima Island.
Recalculate transit costs (shinkansen from Kyoto to Hiroshima instead of Kyoto to Osaka).
Two minutes later, all three files are updated. The budget reflects the different hotel prices and longer train ride. The itinerary has new restaurant recommendations. The packing list added a note about more walking on Miyajima Island.
The plan is not fragile. It is not a house of cards that collapses when you move one piece. It is a set of connected documents that an AI can read, understand, and modify holistically.
Tips for Better Results
Be specific about your travel style. "I like food" produces generic results. "I want one high-end sushi omakase dinner ($80-100/person), otherwise street food and casual ramen joints ($8-15/meal)" produces a plan with real price brackets and a visible contrast between the splurge meal and the daily meals.
Set constraints explicitly. "No taxis," "must be back at the hotel by 9 PM," "one person is vegetarian," "we walk slowly — cap at 15,000 steps per day." Constraints are what make a plan realistic. Without them, you get an ambitious fantasy that falls apart on Day 2.
Use follow-up prompts for depth. After the initial plan:
For Day 4 (Kyoto temples), give me three alternative versions:
1. The popular route (Kinkaku-ji, Fushimi Inari, Arashiyama)
2. The quiet route (lesser-known temples, fewer tourists)
3. The food-focused route (temples near the best lunch spots)
Now you are not just planning. You are exploring options — the way a good travel agent would, except this one works at 2 AM and never gets impatient.
Commit the plan to git. This sounds like a developer move, but it is practical for anyone:
git init
git add .
git commit -m "initial japan itinerary v1"
Now every version of your plan is saved. Changed Osaka to Hiroshima? That is a new commit. Want to go back to the Osaka version? It is still there. Version control for travel plans is not overkill — it is the same as keeping old drafts, except organized.
What This Is Really About
The eight-tab-and-spreadsheet approach fails because it asks you to be the synthesizer. You collect raw data from ten sources and assemble it in your head. That is hard. It is slow. And the result is fragile — change one thing and the whole mental model needs rebuilding.
The terminal approach inverts this. The AI is the synthesizer. You provide the constraints and preferences — the things only you know. The AI handles the assembly, the math, the cross-referencing, and the reformatting when plans change.
You are not removing yourself from the planning. You are removing the drudgery. The choices are still yours — where to go, what to eat, how much to spend. The spreadsheet arithmetic and the tab-juggling are not.
If you have never used an AI CLI tool before, the AI CLI Tools for Non-Engineers guide covers everything from opening a terminal to completing your first project. Start there, then come back here with a destination in mind.
Your next trip is a few prompts away. And for the first time, the planning might actually be fun.