AI for Afghanistan, built and hosted in Kabul.
One API key for chat, translation and summaries that understand Dari and Pashto. Custom AI built by our team. GPU servers to run your own models, with data that stays in the country.
Use our API, let us build it, or run your own
AI API
One endpoint that works with the common chat-completions format. Chat, translate, summarize and read documents, with keys and usage in the portal.
- Dari and Pashto tuned prompts
- Open models hosted in Kabul
- Premium cloud models on the same key
AI development
Our software team builds AI into your work: chat assistants, document reading, call summaries and smart reports, connected to your systems.
- Clickable mockup first, free
- Works with HRM, Finance, CRM, Tickets
- Training in Dari and Pashto
GPU VPS
Virtual servers with dedicated GPU memory in the Kabul data centre. Start from a ready image and run your own models privately.
- Ready images: Ollama, vLLM, Jupyter
- Private link to your Kayanat line
- Hourly or monthly billing in AFN
Drop it into any app in five minutes
If your code already talks to a chat-completions API, change the address and the key. Usage, limits and invoices sit next to your internet bill in the portal.
- Keys per projectSeparate keys, spending caps and alerts by SMS or e-mail.
- Data stays in KabulRequests to open models never leave the country and are not used for training.
- Live usageTokens, cost and errors per key, per day.
# Python · works with the usual OpenAI-style client from openai import OpenAI client = OpenAI(base_url="https://api.kayanat.af/v1", api_key="YOUR_API_KEY") r = client.chat.completions.create( model="kayanat-fast", messages=[{"role": "user", "content": "Translate to Dari: Your line is active."}]) print(r.choices[0].message.content)
| Model | Where | Per 1M tokens |
|---|---|---|
| Kayanat Fast | Kabul GPU | 150 AFN |
| Kayanat Pro | Kabul GPU | 600 AFN |
| Premium cloud | Routed abroad | from 1,800 AFN |
| Document reading (OCR) | Kabul GPU | 5 AFN / page |
What we build with AI
Real projects for ISPs, banks, NGOs and government offices. Every one starts with a free clickable mockup.
Customer chat assistant
Answers customers on your website and WhatsApp in three languages, and opens a ticket when a person is needed.
Document reading
Reads Dari and Pashto forms, IDs and invoices into your system, so staff stop typing them by hand.
Ticket and call summaries
Turns long tickets and calls into a three-line summary with the next step for the team.
Sales assistant in CRM
Drafts quotations and follow-up messages from a customer's history.
Network anomaly alerts
Spots unusual traffic or failing links in NMS data before customers call.
HR helper
Screens CVs against a job, and answers staff questions about leave and payroll rules.
Run your own models on Kabul GPUs
Pick a size, pick a ready image, and connect over SSH or Jupyter in minutes.
Starter
- 1 GPU, 24 GB memory
- 8 vCPU · 32 GB RAM
- 500 GB NVMe
- or 90 AFN per hour
GPU M
- 1 GPU, 48 GB memory
- 16 vCPU · 64 GB RAM
- 1 TB NVMe
- or 190 AFN per hour
Power
- 2 GPUs, 80 GB memory each
- 32 vCPU · 256 GB RAM
- 4 TB NVMe
- Dedicated host on request