An AI workspace that never leaves your building.

Your staff ask questions, draft and summarise in Arabic, English or both in the same sentence. AE-1 runs on your own servers, and works with the internet switched off.

Start a pilotTry it freeHow it works
AE-1 answering live: meeting notes added with the + button become action items, decisions and an Arabic summary, streamed from a model running on our own hardware. Recorded from the product.

Designed in Arabic, not translated into it.

Most tools add Arabic at the end. AE-1 was built right to left from its first line, so mixed-language work reads the way your people write it.

Arabic, English, or both.

The message box follows the language you start in, sentence by sentence, so mixed writing never scrambles.

Six plural forms, counted right.

Arabic changes the noun for none, one, two, a few, many and more. AE-1 follows every rule, in Western or Arabic-Indic digits.

  • Every screen mirrors

    Navigation, menus and icons flip for Arabic. Logos, media controls and code never do.

  • Mixed text stays legible

    Arabic paragraphs with English terms, file names and numbers keep their order; code and links always read left to right.

  • Your digits, your calendar

    Western or Arabic-Indic digits, Gregorian or Hijri dates, set by your organisation and adjustable by each person.

Answers that show their sources.

Bring the documents your teams already rely on. AE-1 answers from them and points to the passage it used, so people can check before they act.

  • Your formats

    PDF, Word, Excel, CSV, plain text and Markdown, in Arabic or English.

  • Search that understands Arabic

    Spelling variants, diacritics and alef forms match the same passage. Exact names and numbers are never lost to paraphrase.

  • Honest about scans

    Scanned pages without a text layer are flagged rather than guessed at.

Easy for IT to run.

One program, one configuration file. Your team can install it in an afternoon on a server with no outside connection.
  • 1binary, with the interface built in
  • 0outbound connections by default
  • 2CPU architectures: x86‑64 and ARM
  • 4themes, in light and dark, each checked for contrast
Install and startshell
$ ae1 config defaults > ae1.toml
$ ae1 --config ae1.toml config check
$ ae1 --config ae1.toml serve
# health: /healthz   readiness: /readyz   metrics: /metrics
Models
Any OpenAI-compatible server you run: vLLM, TGI, llama.cpp, Ollama
Sign-in
Microsoft Entra ID, Okta, Keycloak (OIDC, SAML 2.0, SCIM)
Packaging
Static Linux binaries and a minimal container image
Storage
Embedded database; nothing else to install

Built to pass your security review.

Every access is checked and recorded. Nothing about your conversations is sent anywhere you haven't chosen.

  • No telemetry

    No usage tracking, update checks, remote fonts or third-party calls. It works with all other traffic blocked.

  • Content stays out of logs

    Logs record who, when and which page, never what was asked or answered.

  • Tamper-evident audit

    An append-only, hash-chained record of access that your compliance team can export to its SIEM.

  • Roles by group

    Admins decide which models, connectors and tools each group may use.

Four themes. Both directions.

Sharq, Nord, One and Gruvbox, each in light and dark, every one checked for readable contrast in Arabic and in English. People pick what suits their eyes; your organisation sets the default.

Questions IT and security teams ask.

Does AE-1 need the internet?

No. It only connects to the systems your administrators configure, such as your model server and identity provider. Everything else can be blocked.

Which models does it use?

Yours. AE-1 connects to any model server with an OpenAI-compatible API running inside your network, and admins choose which models each group can use.

Where are conversations stored?

In AE-1’s own database on your server. Admins can set conversations to delete automatically after a chosen number of days.

Can people only see what they’re allowed to?

Yes. Access follows your identity provider’s groups, and documents from connected systems are filtered by each person’s permissions in that system.

Works with

InterstellarConnects AI to your systems, with permissions and an audit trail.Sharq EngineTrains and serves models on your data, on your infrastructure.All productsHow the three work together, on one page.

Start with one workflow.

A pilot runs on your own data, inside your boundary, and is measured against the numbers you already report.

Start a pilotRead the client briefing