✈ TravelAgent Demo

How this demo was made — with two simple prompts

This entire demo — the research briefs, the working agent-portal app, the summary page, and this site — was generated by AI from plain-English prompts. No spec doc, no sprint. That's the point: this demo is itself a demo of how fast new tools for agents and suppliers can now be built.

This is the prompt that generated this site:

“ive built a project plan in project.md. lets build a todo.md for this, and then execute”

One sentence pointing at the project plan — and the AI produced the todo list below, then executed every item: briefs, demo app, summary page, and this site.

Prompt 1 · “Here's my plan — make a todo and execute”

The original one-line prompt referencing project.md
The whole project started with one sentence pointing at project.md.

Prompt 2 · The todo list becomes the build

AI-generated todo list: research briefs, demo app, summary page
The AI turned the plan into a todo list — 4 research briefs, the Python demo portal, the HTML summary — and then executed every item, verifying the app ran before calling it done.

What that proves (for our B2B2C business)

★ Speed: idea → working agent tool in one session, not one quarter.
★ Cheap iteration: supplier spotlight cards, crowd calendars, proposal pages — each is a small build like this one.
★ Agents still matter: AI did the drafting; a human directed, judged, and verified. That's exactly the “AI drafts → agent curates” workflow we're selling.
⚠ Same caveat as the trip plans: AI output needs human verification. We ran the app and checked every endpoint before shipping — do the same for any tool you give agents.

Every source file, readable on this site

Vercel serves static files, so instead of linking raw .md files (which browsers just download as text), each one is wrapped as a page:

The live Python app (app.py) isn't hosted here — its pre-generated output is frozen as the Agent Portal Demo page. Run python app.py locally for the interactive + LLM version.