One API for All Your LLMs

Access Saudi-hosted and global LLMs through a single, secure gateway. Built-in AI safety guardrails, detailed analytics, and complete control.

Quick start
from openai import OpenAI

client = OpenAI(
    base_url="https://dev.alapi.deep.sa/v1",
    api_key="sk-alapi-...",
)

resp = client.chat.completions.create(
    model="deep-sa/alEmbedding",
    messages=[{"role": "user", "content": "مرحبا"}],
)

Drop-in OpenAI-compatible API — change the base URL and you're done.

116,291

Requests served

14

Models

How it works

1

Sign up

Create your account in seconds — no credit card required.

2

Add credits

1 SAR = 10 requests · top up your balance and pay only for what you use.

3

Get your key

Generate an API key and call any model through one endpoint.

Features

Unified Access

Saudi-hosted and global models like GPT-5, Gemini, and more all through one API endpoint.

SafeAI Guardrails

Define safety rules in natural language. Auto-generate prompts and test them before deployment.

Full Visibility

Track every request. Monitor latency, usage, and costs across all providers in one dashboard.

SafeAI — built-in guardrails

Every request can pass through configurable safety guardrails before it reaches a model and before the response returns to your users. Define your rules in plain language and SafeAI enforces them on both prompts and responses — no extra infrastructure to run.

  • Prompt & response analysis

    Inspect both incoming prompts and model output for policy violations.

  • Configurable policies per account

    Set allow and block rules that match your product, tuned and tested from your dashboard.

  • Arabic & English coverage

    Guardrails understand both languages, so nothing slips through the gaps.

AlOCR Document Intelligence

alocr.deep.sa →

Extract text from scanned documents, images, and PDFs with state-of-the-art Arabic and English OCR. Powered by Deep.sa's advanced document processing technology.

Arabic & English text
Automatic deskewing
Layout preservation
Table extraction
Multi-page PDFs
High accuracy

Supported Formats

PDF
PNG
JPEG
TIFF
BMP
WEBP

Max file size: 50MB per document

alRetrieval

Ingest documents (PDF, DOCX, PPTX, TXT, MD) and query them with semantic retrieval. PDFs are processed through alOCR; other formats are extracted directly.

PDF + Text ingestion
DOCX & PPTX support
Semantic search
Auto OCR for PDFs

How it works

  1. 1 Ingest PDFs, DOCX, PPTX, TXT or MD
  2. 2 Documents are parsed, chunked and embedded
  3. 3 Query with semantic search over your collections

Simple Pricing

No token counting. Pay per request - same rate for all models.

LLM Models
1 SAR = 10 requests

Any model, any size input/output

1 SAR

10 req

2 SAR

20 req

4 SAR

40 req

Prices shown exclude 15% VAT

Deep.sa

A product of Deep.sa

Activate account?

Activate this account?