Hackathons: getting and using your key
You are at a hackathon sponsored by AICU. This page is everything you need: how to get your key, one URL, every model on the platform except Anthropic's.
Your key
There are two ways to get a hackathon key. Your organiser tells you which one your event uses.
A. Redeem a voucher on member.aicu.ai (one key per person)
- Open the invitation from your organiser and sign up on member.aicu.ai with the email address your organiser registered. Your AICU ID is created at the same time.
- Go to Redeem and enter the voucher code your organiser gives you.
- Read and accept the AICU Terms of Service and the hackathon programme terms. You cannot redeem without agreeing to both.
- Your sponsored credits and your own key appear. The key value is shown only once — copy it now into a password manager or an environment variable. We do not show it again.
- Later, log in to the dashboard with the same AICU ID. The Hackathon panel shows your credits, their expiry, your key's name and its daily limit (not the key value).
You do not register a payment card. Each person redeems once; if your account already holds AICU credits, use a new account for the hackathon (the two balances cannot expire on different dates).
B. A team key from your organiser
Some events use one key per team. Your organiser (usually a teacher or the event lead) received it and passes it to you. You do not need an account or a card.
Both kinds of key
- The value starts with
aicu_hack_. The name looks likeAICU_Hack_<EVENT>_T<n>(for exampleAICU_Hack_ANGERS_T1), plus a short ID for a personal key. - It expires at most 14 days after it was issued, and it is not extended. After that, every call returns
401withcode: "key_expired". - Keep it secret. Put it in an environment variable, never in a public repository, a web page or an app you ship. A value starting with
aicu_hack_is easy to spot when it leaks. If it leaks or you lose it, email api@aicu.ai right away (and tell your organiser): we stop it and issue a new one.
export AICU_API_KEY=aicu_hack_...
One URL for everything
The API is OpenAI-compatible. Any OpenAI SDK or tool works by changing the base URL and the key.
| Item | Value |
|---|---|
| Base URL | https://api.aicu.ai/v1 |
| Auth header | Authorization: Bearer $AICU_API_KEY |
| Model list (the source of truth) | GET https://api.aicu.ai/v1/models — no key needed |
curl https://api.aicu.ai/v1/chat/completions \
-H "Authorization: Bearer $AICU_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "gpt-5-mini", "messages": [{"role": "user", "content": "Three names for a startup that rents bikes to tourists"}]}'
from openai import OpenAI
client = OpenAI(base_url="https://api.aicu.ai/v1", api_key="aicu_hack_...")
r = client.chat.completions.create(model="gpt-5-mini", messages=[{"role": "user", "content": "Write a one-line pitch"}])
print(r.choices[0].message.content)
What you can build with it
| You want to… | Start with | Guide |
|---|---|---|
| Write copy, a business model, a pitch | gpt-5-mini, gpt-5, deepseek-v3 | Chat (LLM) |
| Write code with an agent | kimi-k2.7-code from OpenCode, Cline or Continue | OpenCode |
| Make a logo or site images | gpt-image-2, nano-banana, sd3.5-large (Stability AI), qwen-image-3.0 | Images |
| Make a short teaser video | wan3.0-video, gemini-omni-1.1-flash | Video |
| Add a voice-over | eleven_v4_turbo, eleven_v4 | Audio |
Prices change, so we do not copy them here. Each model's price is in GET /v1/models and on the pricing page, and every response carries the amount it cost in the X-AICU-AP-Cost header (1 USD = 10,000 AP).
Text, speech and transcription are cheap enough that you will rarely notice them. Images and video are what use up a team budget — a single long, high-resolution video can cost more than everything else your team does in a day. Try a short, low-resolution take first.
What is not available
- Anthropic (Claude) models are not available. There is no
/v1/messagesendpoint, so Claude Code cannot be pointed at this API. For coding, use OpenCode withkimi-k2.7-code. - Google models work, but cost more here than on Google itself. If your team mainly needs Gemini, a personal Google AI Studio or Google Cloud account (with its free tier or trial credit) will go further. Use this key for everything else.
Limits on a hackathon key
They are set when the key is issued. With a personal key (A), log in to the dashboard with your AICU ID to see them in the Hackathon panel. With a team key (B), ask your organiser.
| Limit | What happens | What to do |
|---|---|---|
| Daily cap (always set) | Paid calls are refused until the day resets at midnight Japan time (16:00 CET / 15:00 UTC in winter) | Wait for the reset, or ask your organiser |
| Per-request cap (often set) | 403 request_too_expensive — one call would cost more than the cap | Lower the resolution, the duration or n |
| Allowed models (sometimes set) | 403 model_not_allowed | Pick a model from the allowed list, or ask your organiser |
| Expiry (always set, ≤ 14 days) | 401 key_expired | The hackathon is over — thank you for building with us |
Speech-to-text (
/v1/audio/transcriptions) cannot run on a key with a per-request cap, because its cost is not known before the audio is processed. If your project needs it, ask your organiser for a separate key.
If your team runs out of budget before the end, ask your organiser: more credit can be added to the same key.
Rules
- Use it for the hackathon only. No commercial use, no reselling or lending the key.
- Follow the Terms of Service, including the content rules (no NSFW, nothing illegal).
- Do not publish the key anywhere, including in a demo you share.
When something goes wrong
- Check the error codes — most failures tell you what to change.
- Ask your organiser.
- Use the Support panel on the right of the dashboard (log in with your AICU ID), or email api@aicu.ai. Give the request (endpoint, model, what you sent), the time (with time zone) and the response (status and body). Questions and bug reports in French or English are welcome. A confirmed bug is refunded.
After the event, we share a usage report with your organiser (per team and per kind of output: text, images, speech, video), so you can see what your team actually used.