> For the complete documentation index, see [llms.txt](https://docs.veza.com/4yItIzMvkpAvMVFAamTf/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.veza.com/4yItIzMvkpAvMVFAamTf/features/access-agents.md). # Veza Access Agents AI agents in the Access AI panel that search the Access Graph, explain queries, reviews, and policy runs, troubleshoot integrations, and classify NHIs. {% hint style="warning" %} **AI-generated content:** Access Agents use generative AI. The answers, queries, suggestions, and drafts they return can be inaccurate or incomplete. Review them before acting on them, and read each action card before confirming it. {% endhint %} Veza Access Agents answer natural-language requests about the Veza platform. In Veza, Access Agents open from the **Access AI** panel. Given a request, the panel routes it to the appropriate agent based on the message and the current page. The agent drafts the result and exposes any state-changing operation as an action button, so nothing changes until you select it. See [Access Agents](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents.md) for each agent, what you can ask it, and where it appears in the product. To turn on Access AI and see which agents your tenant has, see [Enabling Access AI](#enabling-access-ai). The agent selector beside the prompt reads **Auto** while routing is automatic. To send a request to a specific agent instead, choose that agent from the selector. Opening a suggested action selects that action's agent. A contextual entry point on a product page, or a suggested action in the panel, starts the appropriate agent with the current page context preloaded. Opening the panel on its own does not start an agent: it returns to the current thread or, in a new chat, shows the suggested actions for the current page. The core Veza platform (Access Graph, Access Intelligence, Access Reviews, Lifecycle Management, NHI Security) operates unchanged with or without Veza Access Agents. Each agent works through the same Veza APIs as the UI and acts under the signed-in user's identity, so team and ownership limits apply. ## Threads Each conversation in the Access AI panel is a thread. Closing the Access AI panel does not end the thread: reopening the panel returns to the same thread. To start a fresh conversation, select **New Chat Thread** in the panel header. Contextual entry points on product pages also open a new thread. To reopen an earlier thread, use **Previous Threads** in the panel header, or **Access AI > Threads** in the main navigation, which lists all of your threads. The options menu (ellipsis icon) on a thread offers **Rename Thread**, **Move Thread** to place it in a folder, and **Delete Thread**. If a response fails, the Access AI panel displays an error in place of that response, with a **Try again** button that runs the request again as a new turn in the same thread. If the connection drops before a response finishes, the panel displays the same error with **Try again**. After you reload the page or reopen the thread, that response shows **Connection lost** with a **Retry now** button. To end a response while it is still generating, select **Stop**; select **Retry** to run the request again. The error does not affect underlying review data, policy state, or graph data; those continue to operate independently of Access AI. ## Enabling Access AI ### Which Access Agents you have Every Veza customer has the [Access Search Agent](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents/access-search-agent.md), the [Integration Agent](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents/integration-agent.md), and the [Veza Documentation Agent](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents/veza-documentation-agent.md). Other Access Agents come with the Veza product they belong to: | Veza product | Additional Access Agents | | -------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Access Reviews | [Access Review Agent](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents/access-review-agent.md) (Early Access) | | Lifecycle Management | [LCM Run Analyst](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents/lcm-run-analyst.md) and [LCM Policy Agent](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents/lcm-policy-agent.md) | | NHI Security | [NHI Owner Suggestion Agent](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents/nhi-owner-suggestion-agent.md), [NHI Classification Agent](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents/nhi-classification-agent.md), and [NHI Agent](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents/nhi-agent.md) | For the availability of other Access Agents, contact your Veza account team. ### What each user needs To use the Access AI panel, each user must meet two conditions: * **Access AI Chat enabled.** Enable **Access AI Chat** in the **Early Access** dialog, which opens from the **Early Access** button in the toolbar. The option is off by default unless your tenant is configured to enable it. The setting is saved in this browser, so another browser starts from the default. If the **Early Access** button or the **Access AI Chat** option is missing, Access AI is not enabled for your tenant. Contact your Veza account team. * **A role that includes Access AI.** The user holds one of these built-in roles: Administrator, Operator, Viewer, Access Reviewer, Access Reviews Admin, Access Reviews Monitor, Watcher, Re-assigner, or NHI Security Admin. See [Roles](/4yItIzMvkpAvMVFAamTf/administration/administration/users/roles.md). Each [agent page](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents.md) lists any gate the agent adds beyond these, such as a data prerequisite. The **Access AI** menu in the main navigation, which contains **Threads**, is shown only to root-team users with the Administrator, Operator, or Viewer role. ## Privacy, data handling, and audit Large language model inference is served through Amazon Bedrock. Agents act under the signed-in user's identity, with the same team and ownership limits and tenant isolation as the rest of Veza. Agent interactions are subject to Veza's data-handling policies, and specifics vary by agent. For environment-specific details on the LLM providers in use, the data included in agent prompts, retention behavior, regional residency, and audit-log integration, contact your Veza account team. ## Related * [Access Agents](/4yItIzMvkpAvMVFAamTf/features/access-agents/agents.md) * [Access Graph](/4yItIzMvkpAvMVFAamTf/features/search/graph.md) * [Access Intelligence](/4yItIzMvkpAvMVFAamTf/features/insights.md) * [Access Reviews](/4yItIzMvkpAvMVFAamTf/features/access-reviews.md) * [Lifecycle Management](/4yItIzMvkpAvMVFAamTf/features/lifecycle-management.md) * [NHI Security](/4yItIzMvkpAvMVFAamTf/features/nhi.md) --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.veza.com/4yItIzMvkpAvMVFAamTf/features/access-agents.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.