The AI Bus is Boarding Soon

Stop Paying Per Token. Start Paying Per Hour.

Flat-Rate AI Compute. Unlimited Tokens API. Paid in Bitcoin.

Bitcoin only
The First time-based AI API
Join The Waitlist

Buy an AI Bus Trip ticket, download the terminal, and board at departure time. For 8 hours, you get a local OpenAI-compatible endpoint. No token counting. No surprise bills.

AI Bus Departures

AI Bus Trips you can take in the next five days. Come ride with us!

🔴 TEST INSTANCE ONLY - NO REAL TICKETS AND API

SUN 25 OCT 2026

— No Departures —

MON 26 OCT 2026

— No Departures —

TUE 27 OCT 2026

— No Departures —

WED 28 OCT 2026

— No Departures —

THU 29 OCT 2026

— No Departures —

Join the Waitlist

Be the first to board when AI Bus Trip launches.

Token-based pricing is broken for agentic coding

Every major API provider — OpenRouter, OpenCode Go, Groq, Claude, GPT, Z.ai, Kimi — charges per token. That model works for chat. It breaks down completely when you’re running agentic coding sessions.

Here’s why:

Qwen models generate a lot of output.
Qwen 27B produces significantly more output tokens than competitors for the same task. One benchmark showed Qwen 3.6 27B burning 98M output tokens where DeepSeek V3.2 finished the same job with 61M. Token-based billing punishes Qwen users especially hard.

Rate limits throttle your agents.
OpenRouter caps you at 600 RPM / 1M TPM. Agentic coding regularly exceeds that. Your agents wait. Your flow breaks. Your bill still climbs.

You never know what it costs.
A “quick 8-hour coding session” can cost $6 or $74 depending on how many tokens your agent generates. There’s no way to budget. No way to predict. No flat rate.

Big context means big bills.
Want 256K context? That’s input tokens on every request. Token pricing means you pay for the same context window over and over again.

We charge for time. Not tokens.

AI Bus Trip runs on a simple model: you buy a ticket for 8 hours. During those 8 hours, you use the GPU as much as you want. No token meter running. No surprise bills. Fair-share scheduling across passengers means your agents always move — even when the bus is full.

The bus metaphor makes it concrete:

  • You buy a ticket for a specific departure (time slot)
  • You board at departure time (Tailscale connect)
  • You sit at your seat (localhost:1234)
  • You ride for 8 hours
  • You get off at the final stop (Tailscale disconnect)

20 passengers share the GPUs. That’s enough capacity for normal agentic coding, chat, and RAG workloads. We show booking numbers transparently — when you see 15 people onboard, you know the performance is real.


Token model charges you for…

AI Bus Trip

Every input token

Nothing

Every output token

Nothing

Large context windows

Included

High request rates

Unlimited

EU Lane Mo-Fr
08:00-16:00 (CET)

Qwen 3.8 27B, FP8, 256k, €9

US Lane Mo-Fr
16:00-00:00 (CET)

Qwen 3.8 27B, FP8, 256k, $9

My Keys To Creating A Powerful Incitin Incident

Let’s make AI API usage affordable. Let’s go on an AI Bus Trip together.

What 8 hours of agentic coding really costs

We compared real-world 8-hour agentic coding sessions on OpenRouter (Qwen3.6-27B at 0.285/2.40 per million tokens + 5.5% fee) vs. the AI Bus Trip flat rate.

Comparison table

Taking an AI Bus Trip saves you up to 90%

Usage scenario

Input tokens

Output tokens

OpenRouter cost

AI Bus Trip

Light user

5M

2M

$6.23

$9

Medium user

15M

8M

$23.48

$9

Heavy agentic

30M

15M

$44.55

$9

Power user

50M

25M

$74.25

$9


OpenRouter costs = (input × $0.285 + output × $2.40) + 5.5% credit fee

Break-even: ~3M output tokens per day

That’s roughly 625K tokens/hour. If you’re doing serious agentic coding with Aider, Hermes, or OpenCode, you’ll hit that in 1-2 hours.

Below 3M output tokens/day → use OpenRouter.
Above 3M output tokens/day → take an AI Bus Trip.

Boarding in three steps

  • Step 1: Buy a ticket

    Choose a lane (EU or US), pick a date, pay €/$ 9 in Bitcoin via BTCPay. You get a boarding code by email. No account, no KYC, no credit card.

    Mark Fields
    PAY WITH BITCOIN
  • Step 2: Download the terminal

    Download the Bus Terminal — a lightweight Mac client that connects you to the GPU server via Tailscale VPN. No Docker, no config, no setup headaches.
    Mary Boyles
    SECURE TAILSCALE CONNECTION
  • Step 3: Board and ride

    At departure time, open the terminal, enter your code, and board. Your seat is ready at http://localhost:1234 — OpenAI-compatible, works with OpenCode.
    Dora King
    http;//localhost:1234
Buy an AI Bus Trip Ticket    →  Download the Terminal    →   Board securely     →   Access the Open AI compatible endpoint      →  Code for 8h
    for €/$ 9                        Mac app                  via Tailscale             @ http://localhost:1234                       Done

What’s Included In Your Ticket

8 hours of GPU time

Flat rate. No token meter. No surprise bills. Use it as much or as little as you want during your ride.

OpenAI-compatible endpoint

http://localhost:1234 on your machine. Tested with OpenCode.

Qwen 27b with 256K context

Full context window on every request. No input token charges. No context compression hacks.

No rate limits

We keep the seats limited per GPUs, so your agent runs at proper speed.

Bitcoin payments

No KYC. No bank. No Stripe. No credit card. Censorship-resistant payments. Your transaction, your business.

Privacy by design

Only email + boarding code. No account at Big Tech. No usage logs. Tailscale VPN encryption. What happens on the bus stays on the bus.

Transparent capacity

20 seats per bus. We show booking numbers. When 15 people are onboard, you know the performance is real — not a marketing claim.

Two lanes, five days

EU Lane: 08:00–16:00 CET, Mon–Fri. US Lane: 16:00–24:00 CET, Mon–Fri. Pick your timezone, board and ride.

Built for people who don’t fit the token model

For developers who hit rate limits

You’re running agentic coding sessions. OpenRouter throttles you. The AI Bus doesn’t; it runs on defined hardware for a limited amount of users.

For Bitcoiners who want to pay in BTC

You don’t want to use Stripe, PayPal, or a credit card. You want censorship-resistant payments. BTCPay, Bitcoin, done.

For AI enthusiasts who want real access

You want Qwen 27B with full 256K context. You want to use it hard for 8 hours. You don’t want to worry about token costs. Flat rate, full access.

For privacy-conscious Coders

You don’t want an account at OpenAI, Google, or Anthropic. You don’t want KYC. You don’t want usage logs. You don’t want your conversations to be trained on. Just an email and a boarding code.

“Taxis charge per kilometer. The bus charges per ride. Same destination, 90% cheaper.“

Louis Thomas

Jonas, Founder AIBusTrip.com

THE FIRST TIME-BASED AI API

Flat-rate compute. Bitcoin only.

We believe Open Source AI Compute should be Simpler

No token meters. Fair GPU share. No KYC. No surprise bills. Just a ticket, a seat, and 8 hours to build.

  • $ 9
  • Bitcoin
  • Done

Questions before boarding?

8 hours of OpenAI-compatible AI compute. No tokens. No rate limits. No KYC. Just €/$9 flat — paid in Bitcoin.

A Mac, Bitcoin, and an email address. No account at Big Tech, no KYC, no credit card.

At departure time. EU Lane boards at 08:00 CET, US Lane at 16:00 CET. Try earlier and you’ll see: “You’re early — your bus arrives at [time].”

The bus departs with a minimum of 15 passengers. If it doesn’t fill, you get a credit for the next ride. No cash refunds — Bitcoin payments are irreversible.

No. Tickets are tied to your email address. Resale is prohibited.

Qwen 27B with 256K context. More “bus lanes” (different models) coming soon.

Bitcoin via BTCPay. No KYC, no bank, no Stripe. Pay from your wallet, get your boarding code by email after 1 confirmation.

After boarding, you get a local endpoint at http://localhost:1234. It’s OpenAI API-compatible — we have tested it with OpenCode, but it should work with Cursor, Continue, Aider, or any tool that supports custom OpenAI endpoints.

Up to 15 passengers share the GPUs. For normal agentic coding, chat, and RAG, performance is solid. We show booking numbers transparently so you know what to expect.

The AI bus is boarding soon

Join the waitlist. Be first to buy a ticket when lanes go live.