Every model. One private chat.

Use Claude, ChatGPT, Gemini and 240+ models side by side. Sensitive data is anonymised before it leaves, and Auto mode picks the right model for each message.

Messages

Auto modepicks for each message
Model Claude And it Thinks longerReads 1 file
  • Claude
  • ChatGPT
  • Gemini
  • DeepSeek
  • Qwen
  • Llama
  • Mistral
  • Falcon
  • K2 Think
  • 240+models
Pick a message, or let it play

Chat with any model, safely.

  • Core

    Anonymised prompts and files

    Use Claude, ChatGPT and Gemini. Names, IDs and secrets are masked first.

  • 240+ models

    Pick a model, compare answers side by side and set your defaults.

  • Auto mode

    Picks the model for each message, and whether to search, think longer or make a file.

  • Bring your own models

    Add custom models, connect your Claude or ChatGPT/Codex subscription, or use provider API keys.

  • Web search and Deep Research

    Multi-step research reports with citations.

  • Memory and instructions

    Custom instructions and memory, private to you.

  • ⌘K

    Global search

    Find chats, files, projects, settings and teammates from anywhere.

  • Folders and smart search

    Sort chats into folders and find them by meaning, not keywords.

  • Knowledge base and projects

    Answers from your own documents, with citations, and reusable project setups.

  • Files, decks and diagrams

    Reports, branded decks, dashboards and diagrams from a single prompt.

240+ models. One platform and API.

The full list of model families, with the newest releases first.

Public AI modelsCloud, anonymised first
  • ClaudeAnthropicClaude Fable 5.2Claude Opus 5.5Claude Sonnet 5.5Claude Haiku 5.5
  • ChatGPTOpenAIGPT-6 AstraGPT-6.1 SolGPT-6.1 TerraGPT-6 Luna
  • GeminiGoogleGemini 4 CarbonGemini 3.8 FlashGemini 3.1 Pro
  • GrokxAIGrok 4.7Grok 4.6
  • MistralMistral AIMistral Large 4Mistral Medium 3.5Mistral Small 4Devstral
Open-source AI modelsRun them on your own servers
  • DeepSeekDeepSeekDeepSeek V4.1 FlashDeepSeek V4 ProDeepSeek V4 Flash
  • QwenAlibabaQwen3.8-MaxQwen3.8 27BQwen3-Coder
  • Llama and MuseMetaMuse Spark OpenLlama 4 MaverickLlama 4 Scout
  • KimiMoonshot AIKimi K3Kimi K2.6
  • GLMZ.aiGLM-5.3GLM-5.2
  • GemmaGoogleGemma 4
  • gpt-ossOpenAIgpt-oss-120bgpt-oss-20b
Built in the UAEArabic-first, sovereign
  • FalconTechnology Innovation InstituteFalcon-H1R 7BFalcon-H1 ArabicFalcon-Emirati
  • K2 and JaisMBZUAI and G42K2 Think V2K2 HorizonJais 3
  • Your own modelsAny OpenAI-compatible or Ollama endpointFine-tuned modelsInternal models
NodeShift One governed gatewayDeploy onAzureAWSGoogle Cloudor your data centre240+ models in total

Questions, answered.

Yes. That is the point. Most regulated teams block these tools because of where the data goes. NodeShift cleans each prompt before it is sent, so you get frontier AI without the exposure.

More than 200 types out of the box, from names, Emirates ID and passport numbers to account numbers, health data and API keys. You can add your own terms and patterns to match your data classification policy.

Yes. Anonymisation keeps the meaning and removes the identifiers. The model gets enough context to do the work, and your team sees the full answer with the real details put back.

Over 240, including Claude, ChatGPT, Gemini, DeepSeek, Qwen, Llama, Mistral and Grok, UAE-built models such as Falcon and K2 Think, and any internal models you already run. Switching is one click, or one parameter in the API.

No. One platform, one key, one bill. You do not manage provider accounts, rate limits or endpoints separately. You can also bring your own keys if you prefer.

Yes. Native Arabic and English, including Gulf dialects, calibrated for legal and regulatory precision. Plus 40+ other languages, and the apps fully support right-to-left text.

Ready to put your AI behind a real firewall?

See NodeShift on your own use cases. A 30-minute live demo, no commitment.

  • UAE PDPL
  • ISO 27001
  • SOC 2 Type II