Not a corporate event. Not a wedding. A regular house party — 25 friends, your apartment, beer and pizza, a Spotify playlist. How much of that can Claude Code handle from scratch?
Inspired by Scott Alexander's SOTA On Bay Area House Party
Let's not overthink this. A normal house party involves roughly these steps:
No event insurance. No liability waivers. No "enterprise venue booking API." Just vibes.
We went step by step and checked what's automatable. Here's the honest answer.
Here's the realistic flow for "Claude Code throws you a house party."
Under the hood, Claude Code orchestrates a handful of services.
$0.20 to text 25 people. Tracks replies automatically. Can also create a Partiful event via their unofficial API.
Creates playlists, controls playback on your speaker, can schedule transitions. The best-automated part of the whole thing.
Instacart for same-day delivery of everything. If API access isn't set up, browser automation on Instacart or any delivery app works fine for a one-off.
No pizza API exists. Browser automation to order, or Fancy Hands ($6) to call it in, or just text you a link to approve.
If you have smart bulbs. Set scenes, schedule transitions, dim as the night goes on. Totally optional — a party works fine with normal lights.
Book a cleaner for day-of and/or next morning. TaskRabbit has a Partner API. Or browser-automate any cleaning platform. ~$100-$150.
Pre-fund a Privacy.com account with $300. Claude gets virtual cards with spending limits. Or just use your saved card on delivery apps and Claude sends you "approve?" links.
Claude Code runs with --dangerously-skip-permissions in a container. MCP servers for Stripe, Gmail, Calendar. Playwright MCP for browser tasks. Hooks for logging.
A regular house party. At your place. 25 friends.
| Item | DIY | Claude Does It |
|---|---|---|
| Beer (2 cases) | $40 | $40 + delivery |
| Wine (2-3 bottles) | $30 | $30 + delivery |
| Mixers / soft drinks | $15 | $15 + delivery |
| Chips, dip, snacks | $25 | $25 + delivery |
| Cups, plates, napkins, ice | $15 | $15 + delivery |
| Pizza (5 large) | $60 | $60 + delivery fee |
| Instacart / delivery fees | $0 | $15-$25 |
| Pre-party cleaning | $0 (your time) | $100-$150 (TaskRabbit) |
| Post-party cleaning | $0 (your hangover) | $100-$150 (TaskRabbit) |
| Spotify Premium | $0 (you have it) | $11 (if not) |
| Twilio texts | $0 (group text) | $0.20 |
| Claude Code API costs | $0 | $5-$15 |
| Total | ~$185 | ~$280-$380 |
The "automation tax" is about $100-$200 — mostly the cleaning services. Whether that's worth it depends on how much you hate cleaning.
Most of a house party is easy to automate. But a few things are real challenges.
This is the fundamental blocker. There's no "social graph API" that tells an AI who you'd want at your party. You have to provide the guest list. But that's a 2-minute copy-paste from your contacts.
DoorDash, Uber Eats, Dominos — none have consumer ordering APIs. This is the biggest gap in the consumer API landscape. Browser automation works (~60-75% reliability) but isn't bulletproof.
It's your apartment and your friends. Someone needs to open the door, point people to the drinks, and be the person whose party this is. An AI can plan the party but can't be the host.
35 states allow alcohol delivery via Instacart/Uber Eats. The others don't. Drizly (the main alcohol delivery API) shut down in 2024. Someone over 21 has to receive the delivery and show ID.
For a regular house party, the honest answer is: Claude Code can handle about 80% of the work, and the remaining 20% takes you about 5 minutes of actual effort.
The interesting finding isn't that AI can plan a party — party planning isn't that hard. The interesting finding is where the consumer API economy breaks down. In 2026, you can programmatically control lights, music, and payments with excellent APIs. But you can't programmatically order a pizza. The "last mile" of consumer commerce — actually buying things on someone's behalf — has almost no API support.
This is changing fast. Visa, Mastercard, and Stripe all launched agent-specific payment products in 2025-2026. Startups like RentAHuman.ai let AI agents hire real humans for physical tasks. But right now, the gap means Claude Code works best as a planner and coordinator that loops you in for the 10-second approval moments.
A house party is the easy case. The research also covers what happens when you scale up.
We did go overboard. 13 parallel research agents, 121 sourced claims, 81 human tasks identified. The full research covers the enterprise case too.