Control agent access tool by tool

Tools are the individual actions inside a provider. Metorial lets you grant or restrict them one at a time. An agent only gets the specific actions it needs.

Get startedBook a demo

Tools

Give agents the exact actions they need

Connecting an agent to a provider shouldn't mean handing it everything that provider can do. Tools are the individual actions inside a provider, and Metorial lets you decide which ones an agent can use. Access lines up with what the agent is actually for.

  1. 01

    Grant access per action.

    Give an agent only the specific tools it needs from a provider, instead of access to the entire integration.

  2. 02

    Separate reading from doing.

    Allow lower-risk actions while blocking sensitive ones. An agent can use a system without being able to make changes you didn't intend.

  3. 03

    Capped by the person behind it.

    An agent can never do more than the person using it. Whatever you grant per tool is still capped by that person's own permissions.

  4. 04

    See every call.

    Each tool call is logged and monitored. You can review which agent used which action and when.

Recent updates to ToolsView all
Custom tool names for Magic MCPBetter Agents, More Models & ToolsIntroducing Metorial – The Future of AI Workflow Building

Pick the exact actions. Grant only what's needed. Block the sensitive ones. Log every call.

FAQ

Answers to common questions about tools and how they fit into governed AI agent infrastructure.

  1. What is a tool in Metorial?
    A tool is an individual action that a provider can perform, such as reading a record or sending a message. Agents call tools to get work done.
  2. Yes. Access can be granted on a per-tool basis. You don't have to give an agent access to an entire provider just to let it use one action.
  3. Yes. You can allow a provider while blocking its sensitive actions, for example allowing an agent to read from Slack but not send messages, through policies and access control.
  4. No. An agent is always bounded by the permissions of the person it acts for. The access you grant per tool can never exceed what that person is allowed to do. See identity and delegation for how this works.
  5. Yes. Every tool call is logged and monitored for security and compliance, and appears in audit logs and observability.
  6. Yes. Tools can be configured with parameters, authentication details, and access control policies so they behave the way each team needs.
  7. A provider is the whole integration, and tools are the individual actions inside it. You manage providers to connect a system, and tools to control exactly what agents can do with it.

Give agents precise access, not broad access

See how Metorial lets you grant and restrict agent access one tool at a time. Book a demo, or get started and connect your first provider.

Book a demoGet started