Publish, discover, govern, host, and serve reusable AI skills, tools, MCPs, models, and agents — securely, sovereignly, and with full UAE Cyber Security Council compliance.
Visible only to the owning individual or entity. For internal agents, work in progress, and Secret or Top-Secret assets.
Shared inside a governed circle of entities — a sector, a consortium, a federal working group — by explicit, audited grant.
Reusable national capabilities any entity can discover and re-serve: Arabic NLP, gov-service templates, common connectors.
A multi-agent council that drafts, reviews, and routes procurement decisions across federal ministries — with a human approval gate on every award and an immutable action audit for non-repudiation.
High-accuracy Arabic optical character recognition for documents, official forms, and handwriting.
Open-weight sovereign large language model tuned for Modern Standard and Gulf Arabic.
Reusable skill that optimizes signal timing from live mobility telemetry across a city grid.
Robustness, adversarial, and safety evaluation harness required before any agent is served to government.
Open-weight multilingual large language model from TII (Abu Dhabi), with vision-language support.
State-of-the-art Arabic–English bilingual LLM (Inception · MBZUAI · Cerebras).
Strong multilingual open-weight LLM with solid Arabic and long-context support.
Multilingual automatic speech recognition and transcription, including Arabic.
No Language Left Behind: machine translation across 200 languages incl. Arabic.
Multilingual, multi-granularity text embeddings for semantic search and RAG.
Transformer language model pre-trained for Modern Standard Arabic understanding.
Arabic NLP toolkit: morphology, dialect identification, NER, and sentiment.
Open-source optical character recognition supporting Arabic script and 100+ languages.
Detect, classify, and redact PII (incl. Emirates ID, IBAN) in text and images.
The standard library for downloading, running, and fine-tuning open models.
Framework for building LLM applications, agents, and retrieval pipelines.
High-throughput, memory-efficient inference and serving engine for LLMs.
Run open models locally with a simple, sovereign-friendly runtime.
Model Context Protocol server exposing scoped, governed filesystem access to agents.
MCP connector for governed, rate-limited HTTP fetch and web content extraction.