The AI models and apps Unseen works with
Approve the providers and models you want, connect the apps people work in, and keep one set of rules across all of them.
Model providers
The providers Unseen connects to in Secure AI Chat and the AI Gateway. You decide which models people may use.
| Provider | What you get |
|---|---|
| OpenAI | GPT models, image, speech |
| Anthropic | Claude models |
| Google Gemini | Gemini models |
| Google Vertex AI | Gemini and partner models in your Google Cloud |
| Azure OpenAI and Azure AI Foundry | OpenAI and other models in your Azure tenant |
| AWS Bedrock | Amazon Nova, Mistral and other models in your AWS account |
| Mistral | Mistral models |
| xAI | Grok models |
| DeepSeek | DeepSeek models |
| Perplexity | Search-grounded models |
| Cohere | Command and embedding models |
| Stability AI | Image generation |
| Your own models | Any OpenAI-compatible endpoint, including models you host |
Connected apps
Apps people can use from Secure AI Chat and agents can reach through the gateway. Each tool can be allowed, held for approval or blocked.
- Microsoft 365
- GitHub
- GitLab
- Atlassian (Jira and Confluence)
- Azure DevOps
- Slack
- Salesforce
- Notion
- Linear
- Asana
- Intercom
- Sentry
- Gamma
- DeepL
How tool rules work: AI Agent Security.
Where people use it
The same models and the same rules, wherever the AI is used.
Secure AI Chat
The chat workspace, with projects, knowledge bases, agents and side-by-side model comparison.
Word, Excel and Outlook
The same models and rules inside the Office apps people already use.
AI Gateway
An OpenAI-compatible endpoint for your apps and agents, including MCP tools and A2A agents.
And the AI tools you did not approve
ChatGPT, Microsoft Copilot, Claude, Gemini, GitHub Copilot, Perplexity and the long tail of smaller tools are services Unseen governs rather than serves: it shows who uses them and applies your rules. The AI Service Directory lists their data terms and retention.
Questions
Which AI models does Unseen support?
Models from OpenAI, Anthropic, Google Gemini and Vertex AI, Azure OpenAI and Azure AI Foundry, AWS Bedrock, Mistral, xAI, DeepSeek, Perplexity, Cohere and Stability AI, plus any OpenAI-compatible endpoint, including models you host yourself. You choose which ones people may use.
Does Unseen work with Microsoft Copilot, ChatGPT and other AI tools people already use?
Yes, but in a different way. Tools such as ChatGPT, Microsoft Copilot, Claude, Gemini, GitHub Copilot and Perplexity are services Unseen governs: it sees their use and applies your rules. The models above are the ones Unseen serves in Secure AI Chat and the AI Gateway.
Do the providers train on our data?
Unseen reaches models through API access that excludes model training, on provider accounts you choose. With Azure, AWS Bedrock, Vertex AI or your own models, the models run in your own cloud account.
Can people choose the model?
Yes, from the models you approve. Admins decide which models people may use, and people can compare answers from several models side by side.