Skip to content
Access
Request access

Serverless APIs

AI in your app, one call away.

Call hosted open models for text, speech and vision through a simple API. No GPUs to size, start or stop — and every request is processed in Nepal.

Request accessHow it works

Who it's for

Adding AI features to apps and websites.

  • App and web teams

    Add a chat assistant, a summary button or a Nepali voice feature without hiring an AI team.

  • Banks and insurers

    Read forms, classify messages and draft replies, with customer data processed inside Nepal.

  • Startups

    Try an idea in an afternoon and pay only for the requests you make.

What teams use it for

  • A customer-service assistant that answers in Nepali and English
  • Summaries of long reports and loan files
  • Turning scanned forms into structured data
  • Speech to text for call-centre recordings

How it works

Three steps.

  1. 01

    Get an API key

    From your workspace, once access is open.

  2. 02

    Pick a model

    Open models for text, speech and vision, hosted and ready.

  3. 03

    Call it from your app

    With the OpenAI-style client code you may already have.

What you get

Serverless APIs, in detail.

  • No GPUs to manage

    We keep the models running and scaled. You send a request and get an answer.

  • Familiar API format

    Language-model endpoints follow the OpenAI API format, so existing client libraries carry over.

  • Text, speech and vision

    Chat, summarise and classify text, transcribe speech, and read images and documents.

  • Processed in Nepal

    Prompts and answers are handled on GPUs in our Kathmandu Valley datacentre, never sent abroad.

  • Pay for what you call

    Usage-based billing in rupees, with no reservation and no minimum term.

  • Grow into your own model

    When you need a model of your own, move to managed inference on the same platform.

Under the hood

The key numbers.

Location
Kathmandu Valley, Nepal
Round trip inside the Kathmandu Valley
< 5 ms
Currency
Nepalese rupees (NPR)
Full specifications

Billing

By usage, in NPR

Commitment: none. Rates are being finalised; every request gets a firm quote in rupees.

How billing works

Questions

About serverless apis.

Which models are available?

A selection of popular open models for text, speech and vision. Tell us what you need and we confirm it is available.

Do our prompts leave Nepal?

No. Requests are processed on GPUs in our datacentre in the Kathmandu Valley.

Do you train on our requests?

No. We never train anything on your prompts or answers.

Start with serverless apis.

Tell us what you want to run. It takes about a minute.

Request access

Other ways to run