AI assistant & API
A fast AI assistant — and an API you can ship on.
Streaming answers, conversation history, documents, vision and web search — with an OpenAI-compatible API for builders, in English and Arabic.
Built to do real work
Not a demo — the kind of assistant you keep open all day, and an API with the billing discipline of a real platform.
An assistant with memory and manners
Conversations stream in real time and persist across sessions — pick up where you left off, edit prompts, branch threads. Drop in screenshots, PDFs, spreadsheets or code and ask about them.
- ✓ Streaming chat with conversation history
- ✓ Document generation — download as PDF, Word or Markdown
- ✓ Vision: screenshots, charts, scanned documents
- ✓ Web search with cited live sources
Drop-in OpenAI-compatible API
Point any OpenAI SDK at ai.pearlfibers.com/v1, swap your key, ship. Streaming, usage headers and error shapes all match what your code already expects.
- ✓ Standard /v1/chat/completions
- ✓ All three models, streaming and non-streaming
- ✓ Quickstart in curl, Python and JavaScript
Billing you can put in a budget
Pay per token in micro-dollar increments — no subscriptions, no minimums. Rates are frozen on each request, so a price change never rewrites a historical bill.
- ✓ Per-key rate limits and monthly caps
- ✓ Usage dashboard and audit log
- ✓ Costs computed per request, frozen at write time
Arabic as a first language
The assistant answers in Arabic as naturally as English, and the whole product — chat, settings, docs, emails — ships fully bilingual with true right-to-left layout.
- ✓ Bilingual chat, interface and documentation
- ✓ True RTL experience, not a mirrored afterthought
Your data stays put
We do not train on your prompts. Peak AI is built and operated by PearlFibers — an Iraq-based product team that uses it every day.
- ✓ No training on customer prompts
- ✓ Operated on infrastructure we run ourselves
Who Peak AI is for
One assistant for the team, one endpoint for the product.
Teams that live in an assistant
Drafts, summaries, analysis, documents — streaming answers with history, all day, in either language.
Builders shipping AI features
An OpenAI-compatible endpoint means your existing SDKs, retries and telemetry work unchanged — swap the base URL and go.
Arabic-first organizations
An assistant that is genuinely bilingual — not translated marketing over an English-only product.
Free to chat, pay-as-you-go to build
Chat is free with a PearlFibers SSO account. The API is pay-per-token with published per-million-token rates — billed in tiny increments, with no subscriptions, no minimums and no surprises.
- ✓ Chat: free with your PearlFibers account
- ✓ API: pay-as-you-go per million tokens
- ✓ Request API access from Settings → API — approved within 24 h
Current per-model rates are published on ai.pearlfibers.com/pricing and frozen per request at billing time.
Questions, answered
How do I get API access?
Sign in to Peak AI, open Settings → API and request access — an admin reviews and approves within 24 hours. Keys come with per-key rate limits and monthly caps you can see at all times.
Is it really OpenAI-compatible?
Yes — point any OpenAI SDK at https://ai.pearlfibers.com/v1 with a Peak key. Chat completions, streaming, usage headers and error shapes match the format your code already handles.
Do you train on our prompts?
No. Your prompts and files are used to answer you, not to train models.
Which model should I use?
Peak AI V4.3 for deep reasoning, code and analysis; V4.3 Light as the fast, cheap default for chat and drafts; V4.3 Vision when images are involved. All three are available in chat and through the API.
Start using Peak AI in seconds.
Sign in with your PearlFibers account and ask for something real — or wire the API into your product this afternoon.