> For the complete documentation index, see [llms.txt](https://docs.zerve.ai/zerve-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.zerve.ai/zerve-docs/security/agent-approvals-and-governance.md). # Agent approvals and governance How Zerve's agents ask for approval, and the controls you have over what they do. Zerve's agents can do real work: run code, deploy apps and use tools that reach other systems. This page explains when they ask for approval, and the controls you have over what they do. ## What agents ask before doing | Action | What happens | | --------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Deploying** | Before starting, updating or stopping a preview or production deployment, the agent asks you to confirm. Each user can turn this off with **Don't ask before agent deployments** in [Preferences](/zerve-docs/administration/preferences-and-notifications.md#agent). | | **Using a tool that requires approval** | For [MCP servers](/zerve-docs/agents-in-the-workspace/mcp-servers.md) marked **Requires approval**, the agent asks before using any of their tools | | **Irreversible changes** | The agent always waits for you before doing something that can't be undone, whatever its other settings | | **Using a credential it doesn't have** | The agent always waits for you to provide it | {% hint style="warning" %} **Superagent workers don't ask before deploying.** They work in the background, where there's nobody to answer, so deployment approvals are skipped for them. Use a superagent's [guardrails](/zerve-docs/superagents/settings.md#persona-and-guardrails) to set what it may and may not deploy, and an objective's [autonomy level](/zerve-docs/superagents/objectives-kpis-and-autonomy.md#autonomy) to have the superagent ask you first. {% endhint %} ## Controls for Workspace agents | Control | What it does | | --------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------- | | [**Agent modes**](/zerve-docs/agents-in-the-workspace/agent-modes.md) | Ask mode can't change anything. Plan mode makes changes only after you approve its plan. | | [**When the agent needs input**](/zerve-docs/agents-in-the-workspace/evolve-mode/when-the-agent-needs-input.md) | Whether Evolve Mode waits for your answers or carries on | | [**Credit caps**](/zerve-docs/agents-in-the-workspace/evolve-mode/credit-caps.md) | The most a single Evolve Mode run can spend | | [**Custom subagents' tools**](/zerve-docs/agents-in-the-workspace/custom-subagents.md) | Each custom subagent can only use the tools you give it | | [**MCP tools**](/zerve-docs/agents-in-the-workspace/mcp-servers.md) | Which tools from each server the agent may use | People with view-only access to a notebook can only use Ask mode there. ## Controls for superagents | Control | What it does | | --------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------- | | [**Guardrails**](/zerve-docs/superagents/settings.md#persona-and-guardrails) | Rules the superagent follows strictly. They can't override Zerve's own safety rules. | | [**Autonomy levels**](/zerve-docs/superagents/objectives-kpis-and-autonomy.md#autonomy) | Per objective: how freely the superagent decides, from **Ask me first** to **Full autonomy** | | [**Decisions**](/zerve-docs/superagents/decisions-and-recommendations.md) | The superagent raises a decision instead of guessing, including before irreversible changes | | [**Autonomous mode**](/zerve-docs/superagents/settings.md#autonomous-mode) | Whether the superagent keeps working through its backlog on its own. Off by default. | | [**Daily budget**](/zerve-docs/superagents/budgets-and-credits.md) | The most its whole team can spend in a day | | [**Channel access**](/zerve-docs/superagents/channels.md) | Who can message the superagent from outside Zerve. Closed by default. | | [**Pause**](/zerve-docs/superagents/monitor-a-superagent.md#pause-resume-and-delete) | Stops all of the superagent's work immediately | People with view-only access to a superagent can ask it questions, but can't start work or change anything. ## Seeing what agents did Every agent's work happens in a notebook, with real code and saved results, so you can check exactly what was done: * A superagent's **Activity** feed records what each team member did. See [Monitor a superagent](/zerve-docs/superagents/monitor-a-superagent.md#activity). * Every model call is recorded, with the model used and its cost. See [Usage monitoring](/zerve-docs/administration/usage-monitoring.md). * Superagents record their decisions and where each came from in their [memory](/zerve-docs/superagents/memory.md). ## Organization-wide controls | Control | What it does | | -------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------- | | [**Model settings**](/zerve-docs/administration/model-settings.md) | Which model providers and models anyone can use | | [**Member permissions**](/zerve-docs/administration/members-and-roles.md#member-permissions) | Which features each member can use, including superagents and deployments | | [**Block restrictions**](/zerve-docs/administration/organization-settings.md) | Lets admins restrict what blocks can do | --- # 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.zerve.ai/zerve-docs/security/agent-approvals-and-governance.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.