Do not use a microscope to hammer a nail.
Most of what your app needs from AI isn't frontier intelligence — smaller tasks deserve smaller models, at a fraction of the price.
Get an API key
Sign up free — no credit card required. Your key is ready in seconds.
POST your task
task says what to do, input is your content. Any language, no SDK.
Use the output
Read output from the response — plain text or JSON, ready to use.
What teams build with it
urgent!! order didnt arrive
Hi,
urgent!! order #48291 didnt arrive and i need it for a client presentation tomorrow at 9am. tracking says "in transit" since thursday which makes no sense.
please refund or fix this asap!!
thanks
Marco
{
"intent": "status_update",
"urgency": "critical",
"order_id": "48291",
"action": "escalate to carrier, contact before 9am"
} → routed to support-urgent Contact forms · Support routing · Newsletter triage
"Hallo, ich werde doppelt abgerechnet – einmal für den alten Plan, einmal für den neuen. Bitte sofort korrigieren."
09:47"Hello, I am being charged twice — once for the old plan and once for the new. Please correct this immediately."
{
"intent": "billing_dispute",
"urgency": "high",
"category": "billing"
} Multilingual support · Global feedback · International e-commerce
12.1 Processor shall process personal data solely on documented instructions of the Controller, unless required to do so by applicable law. Processing shall be limited to EEA territory except where explicitly agreed...
Provider shall give Customer no less than 30 days prior written notice before engaging any new subprocessor. Customer may object in writing within 30 days of receipt. Upon a valid objection, either party may terminate...
{
"data_region": "EEA only (§12.1)",
"notice_period": "30 days",
"exit_window": "60 days, no penalty",
"auto_renewal": true
} 50-page contracts · Dense reports · No chunking, no preprocessing
"I'll handle the webhook — done by end of week."
"I need to send the API credentials — that's blocking us."
"Sarah: I'll handle the webhook by end of week. Marcus: I need to send API credentials — that's blocking us right now..."
[
{"task":"Webhook setup","owner":"Sarah","due":"Fri"},
{"task":"Send API creds","owner":"Marcus","blocking":true}
] Sales calls · Customer interviews · Internal standups
Only Instruct LLM model available for Beta · Coming soon: Embeddings · Audio · Vision · Reasoning · and more
Common questions
Not finding what you need? Questions about pricing, limits, or your specific use case — we reply fast.
[email protected]Which model does it use?
A smaller, faster instruct model — purpose-built for tasks like classification, extraction, and summarisation. You don't pick the model; we match one to your task automatically. No bloated frontier model where a lighter one gets the job done.
Is my data stored or used to train models?
No to both. Your input is discarded the moment the response is returned — nothing is logged, nothing is retained. Your data is never used to train or improve any model, on any plan, ever.
What format does the response come back in?
Whatever your task prompt asks for. Request JSON and you get JSON. Ask for a plain sentence and you get a plain sentence. There's no fixed wrapper or extra parsing step — the response body is the model's output, ready to use.
How is this different from calling OpenAI directly?
Three things: smaller models (faster, cheaper, right-sized for the task), flat monthly pricing (no token counting, no end-of-month surprises), and zero configuration (no model to pick, no parameters to tune — just POST your task). Using GPT-4o to classify a support ticket costs roughly 10× more, and you still have to choose a model, set parameters, and count tokens.
Where is my data processed? How do you handle privacy?
Our API and orchestration run in the EU. GPU inference may use compute outside the EU — but your input is never written to disk at any point in the process. It's handled in memory and discarded the moment the response is sent. No storage, no logging, no training on your data, anywhere.
Are there rate limits?
Yes, but they're a soft guardrail against abuse, not a hard ceiling on what the API can handle. If you exceed the limit you'll get a 429 — back off briefly and retry. Exact limits by plan are in the docs. Need higher throughput? Get in touch.
What's coming after beta?
Right now, a8k.me runs a single model type — an Instruct model, purpose-built for tasks like classification, extraction, and summarisation.
After beta, two things are coming:
- More model types. Embeddings, Vision, Audio, and Reasoning — each purpose-built for its task, all under the same flat-rate model.
- A larger, more capable Instruct model. The current one handles most jobs well. After beta, we'll upgrade to a bigger model that handles longer and more complex jobs.
When those ship, Solo moves to $8.99/month — still one flat price, all models included, no token counting. You'll hear from us before anything changes.
We're also building a Team plan — designed for SaaS products and production apps that need higher throughput, higher limits, and the kind of reliability you can build a business on. Think unlimited-feel usage, higher concurrency, and room to scale without watching a meter.
Do I need a specific SDK or language?
No. It's a plain HTTP API — any language or tool that can make a POST request works. Python, Node, Go, Ruby, curl. If it speaks HTTP, it works.