> ## Documentation Index > Fetch the complete documentation index at: https://dev-docs.altrum.ai/llms.txt > Use this file to discover all available pages before exploring further. # AI systems > Register the AI systems your organization builds or buys, keep their details current, and switch them on and off. An AI system is something your organization builds or buys that uses a model: a support assistant, a fraud scorer, a drafting tool. **AI systems** in the sidebar is your register, with one record for each. Every member can see the register. Owners and admins add, edit, activate and deactivate AI systems. ## Add an AI system Open **AI systems** and choose **Add AI system**. | Field | Rules | | - | - | | Name | Required, up to 200 characters. Unique in your organization, ignoring capital letters. | | Description | Required, up to 2000 characters. What the AI system is. | | Purpose | Required, up to 2000 characters. What it is for and who uses it. | | Owner | Required. An active member of your organization, you by default. The owner can run its assessment and configure its guardrails whatever their role. | | Initial risk | Low, Medium, High or Very high. Your own first view, before any assessment. | | Providers | At least one: OpenAI, Azure OpenAI, Azure AI, Anthropic, Bedrock, Google Gemini, Google Vertex AI. | Saving takes a few seconds: keep the window open. A new AI system is **Inactive**. You can add up to 60 AI systems an hour. If the name is taken, you see "Your organization already has an AI system with that name. Choose a different name." ## The AI system's page Select an AI system in the list to open it. The top of the page shows its details and where its assessment stands: | Assessment state | Button | | - | - | | Not assessed yet | **Start assessment** | | Assessment in progress | **Resume assessment** | | Assessment ready for review | **Review assessment** | | Last assessed on a date | **View results** and **Start new assessment** | When someone else is working on the assessment, the page names them. Below it are tabs: | Tab | What it shows | | - | - | | **Guardrails** | The live guardrails: each guardrail, what it applies to, what it does when it fires and what it looks for. A badge says **Live: version N** or **Nothing is live**, and **Changes saved, not applied** when the draft differs. A link opens the guardrail editor. | | **Risks and controls** | The risks and controls from the latest submitted assessment, read only. | | **AI Gateway** | Owners and admins, when your organization has the AI Gateway. **Keys**: the AI system's keys, with **Create key** to make one on it. A key shows **Unguarded** while the AI system has no live guardrails, and a notice says so. **Limits**: the budget and rate limit across every key of the AI system. **MCP tools**: which MCP servers and tools every key of the AI system reaches (see [MCP tools](/ai-gateway/mcp#choose-who-can-use-which-tools)). **Cache**: **Follow the organization**, **On** or **Off**, and **Clear cache**. **Content logging**: whether its prompts and replies are kept, **Follow the organization**, **On** or **Off** (see [Request logs](/ai-gateway/request-logs#content-is-off-by-default)). The AI Gateway is coming soon. See [AI Gateway](/ai-gateway/overview). | | **Assessment history** | Every assessment: when it started, who started it, its status and when it completed. | | **Overview** | Coming soon. | | **Cost and usage** | Coming soon. | ## Edit an AI system Owners and admins open the menu next to **Activate** or **Deactivate** and choose **Edit**. Every field can change, including the owner. A name already in use is refused and nothing changes. ## Activate and deactivate **Activate** switches the AI system on, and **Deactivate** switches it off. Each asks you to confirm and says what will happen, including whether guardrails are live: * Activating: "Applications can reach it through the AI Gateway again. Owners are told by email." * Deactivating: "Applications that call it are refused until you activate it. Owners are told by email." The AI system's owner and the owners of your organization are emailed. While an AI system is inactive, its page says "This AI system is inactive. Requests through the AI Gateway are refused." You can switch it on and off as often as you need. Assessments and guardrails work whether it is active or not. If you see "This AI system is still being set up. Try again in a few minutes.", Altrum is still finishing the AI system you just added. Nothing is lost: try again shortly. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.