> ## Documentation Index > Fetch the complete documentation index at: https://docs.nexalytica.com/llms.txt > Use this file to discover all available pages before exploring further. # Configure AI Models, Agents, and Tools for Your Org > Control which AI models, agents, and tools are available across your organisation. Set defaults, suspend resources, and manage memory and skills. The **AI Capabilities** section gives org admins precise control over every building block that Nexalytica agents rely on. From choosing which language models are available to approving the tools agents can call, every decision you make here shapes what agents can and cannot do across the entire organisation. Navigate to **Settings → AI Capabilities** to get started; your last-visited tab is remembered in the URL. All settings in AI Capabilities are **org-admin only**. Regular members, department managers, and employees work within the models, agents, and tools you approve here — they cannot access this section. *** ## AI Models The **AI Models** tab lets you decide which language models agents may use and which model acts as the organisation-wide default fallback. ### Reading a model card Each enabled model appears as a card showing the information you need to make cost and capability decisions. | Field | Description | | ----------------------- | ------------------------------------------------------------------------------ | | Provider badge | Identifies the model's provider (e.g. OpenAI, Anthropic, Google) | | Model name | Display name of the model | | Context window | Maximum token context the model supports | | Input cost / 1K tokens | Cost charged per 1,000 input tokens | | Output cost / 1K tokens | Cost charged per 1,000 output tokens | | Capability tags | **Tools** (function/tool calling) and/or **Streaming** (token-by-token output) | ### Default and suspended states | Badge | Meaning | | --------------- | --------------------------------------------------------------------------- | | Crown (Default) | This model is the org-wide fallback when an agent has no specific model set | | Suspended | The model is turned off for the whole org; the card appears dimmed | ### Managing models Click **Set Default** on any active model card to make it the organisation's fallback model. The previous default loses its crown badge immediately. Any agent that hasn't been configured with a specific model will start using the new default from that point forward. Choose a default model that balances cost and capability. Agents that rely on the default inherit whatever model you set here, so a change affects every unconfigured agent at once. Click **Suspend** on a model card to turn it off for the whole organisation. Suspending a model stops every agent that uses it. A confirmation dialog will appear and ask you to type a safety phrase before the action completes. Affected agents automatically fall back to the org default model — verify your default is suitable before suspending. Once a model is suspended its card dims and a **Suspended** badge appears. To see all suspended models, enable the **Show suspended (N)** toggle — this toggle only appears when at least one model is suspended. Click **Restore** on any suspended model card to bring it back online. Agents previously using that model resume using it immediately; no additional configuration is required. ### Enabling a new model from the catalog Click **Enable New Model** to open the model catalog dialog and add a model that isn't yet in your enabled set. Use the **search box** to filter by model name, provider, or model ID. Use the **All providers** multi-select to narrow to specific providers — each provider shows how many models it has in the catalog. Use the **All features** filter to show only models that support **Tool Calling**, **Streaming**, or both. Click any column header — **Model**, **Provider**, **Context**, **Input/1K**, or **Output/1K** — to sort the catalog by that field. Click **Enable** next to the model you want to add. Models you have already enabled show an **Enabled** tag instead of a button. The row counter at the bottom of the dialog shows how many models are listed, the catalog total, and how many are already enabled. *** ## Agents The **Agents** tab is your central hub for browsing, finding, and managing every agent in the organisation, and for setting which agent each department uses by default for each operating mode. ### Finding agents Use the **search bar** to filter agents by name, description, or department. Use the **All departments** multi-select to narrow results to one or more departments. Enable the **Show suspended (N)** toggle to include suspended agents in the list. ### Agent cards Each agent card surfaces the key details at a glance. | Element | Description | | ------------ | ----------------------------------------------------------------------------------------------------------- | | Name | Agent's display name | | Departments | Departments the agent is assigned to | | Description | Short summary of what the agent does | | System badge | Marks built-in Nexalytica agents that cannot be deleted or suspended | | Status dot | **Green** = active, **Amber** = inactive, **Red** = suspended | | Mode chips | Shows which of the 5 operating modes (Quick, Deep, ReAct, LangGraph, Persistent) are enabled for this agent | ### Actions on an agent card | Action | What it does | | ------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Configure** | Opens the agent's mode-configuration editor — a tab per execution mode (Quick, ReAct, LangGraph, Deep, Persistent) where you set that mode's model, tools, prompts, and settings. Changes auto-save per mode. | | **Duplicate** | *Coming soon* — not yet available | | **Suspend** (power icon) | Turns the agent off for the whole org; requires typing a confirmation phrase. Not shown for System agents | | **Restore** (power icon) | Re-activates a suspended agent | | **Delete** | *Coming soon* — currently shown only on suspended non-system agents | ### Default agents per department The **Default Agents** section at the top of the Agents tab shows one row per department. For each department you can pick a default agent for all five operating modes: Fast, low-latency responses for simple queries Extended reasoning for complex analytical tasks Reason-then-act loop for multi-step problem solving Graph-based workflow orchestration Long-running agents that maintain state across sessions Use the picker in each cell to assign the agent your department should use when a user doesn't select one explicitly. ### Creating an agent Click **Create Agent** to start building a new agent from scratch. The creation flow opens Agent Foundry where you configure the agent's name, departments, operating modes, model, tools, and system prompt. *** ## Tools The **Tools** tab lets you approve which tools agents are allowed to call. Tools are organised in a three-level tree: **tier → category → tool**. Use the checkboxes at any level to enable or disable an entire tier, category, or individual tool in one click. ### Tool tiers Always enabled and locked. You cannot turn system tools off — they underpin core agent functionality. Pre-enabled when you set up your org. You can turn these off if your org doesn't need them. Off by default. Turn these on explicitly when your org needs additional capabilities. Some tools carry additional labels: | Label | Meaning | | --------------- | ----------------------------------------------- | | **Preview** | Tool is available but produces limited output | | **Coming soon** | Checkbox is disabled; tool is not yet available | ### Enabled tools summary A live chip strip at the top of the Tools tab shows every currently approved tool, grouped by category. Click the **×** on any chip to remove that tool from your draft selection without scrolling to find it in the tree. ### Saving changes The page tracks unsaved edits and displays either **"You have unsaved changes"** or **"All changes saved"** so you always know whether your draft has been applied. Click **Save** to apply your changes. If your save removes any previously approved tools, Nexalytica shows a confirmation dialog before proceeding. **Removing approved tools** also removes them from every agent mode configuration that was using them. The **Remove approved tools?** dialog shows your current and new approved-tool plan side by side. Click **Confirm & save** to proceed or **Cancel** to go back and review. After saving, a summary reports how many tools were added, how many were removed, and how many agent mode configurations were automatically updated. *** ## Memory The **Memory** tab lets you manage what agents learn and remember across interactions. Nexalytica's memory layer is powered by **Zep Cloud**. Use the sub-sections below to monitor memory health, review pending facts, and maintain compliance. A dashboard showing overall memory health and recent activity across your organisation. Use this as your starting point to spot anomalies before drilling into specific queues. Facts that agents have surfaced and are waiting to be promoted into longer-term memory. Review the queue to ensure that only accurate, relevant information advances. Proposed memory or behaviour rules that agents have inferred, awaiting your review and approval. Approve candidates to encode them as persistent rules, or dismiss them to prevent agents from acting on them. Search and inspect individual memory entries stored in the knowledge graph. Use this to verify what agents know, investigate unexpected agent behaviour, or audit specific facts before a data-removal request. A record of every memory change and action, including what was added, modified, or removed and by which agent or user. Use this log for compliance review and incident investigation. Privacy controls for handling personal data held in memory. Use this section to locate personal data, fulfil data-subject access requests, and remove PII from the knowledge graph to meet your GDPR obligations. Click **Export audit** at the top of the Memory tab to download a full audit export of memory activity for offline review or compliance records. *** ## Skills (Preview) **Skills is a preview feature showing sample (mock) data only.** Nothing on this tab affects your agents yet. It is provided so you can explore the concept and give feedback before the feature reaches general availability. Skills are multi-step tool sequences that an agent can execute as a single named action — for example, a "Monthly Report" skill might pull data, run analysis, and draft a summary in one invocation. The tab displays sample skill cards, each showing: | Element | Description | | -------------- | ------------------------------------------------------------------------------- | | Name | The skill's display name | | Description | What the skill does | | Pipeline steps | The sequence of tools the skill orchestrates | | On/Off switch | Display only — switches are disabled in the preview and do not control anything | Check back when Skills reaches general availability to start building and enabling real skill pipelines for your agents.