How-to

Manage agent access and visibility

Teach admins to review which knowledge an agent can reach. An agent should receive the context it is permitted to use—not everything.

1 min 10 sFor: Workspace admin9 steps

Read the steps OR watch the video. Both teach the same outcome, and neither is needed for the other.Jump to the video

Quick steps

  1. Click SettingsAgents are managed from Settings. Select Settings in the sidebar.
    1. Start on DM Home (/).
    2. Click "Settings". The next view is "Workspace settings" (/settings/workspace).

    What it means: Agents are managed from Settings. Select Settings in the sidebar.

  2. Click AgentsSelect Agents.
    1. Click "Agents". The next view is "Agents" (/control-panel/agents).

    What it means: Select Agents.

  3. Click Select agent (Claude Code)This workspace has one agent, Claude Code. Open it to see what it is allowed to do.
    1. Click "Select agent (Claude Code)". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001).

    What it means: This workspace has one agent, Claude Code. Open it to see what it is allowed to do.

  4. Review capabilities enabledThe Access step summarizes how many capabilities this agent has enabled.
    1. Review "Capabilities enabled": "Access 12 capabilities enabled Governed capabilities this Agent may use. View".

    What it means: The Access step summarizes how many capabilities this agent has enabled.

  5. Review current profileUnder Agent access, an access profile sets the agent's baseline. Standard is selected. It reads authorized knowledge and can submit proposed changes for review, but has no approval authority.
    1. Review "Current profile": "Standard Read access to authorized shared knowledge; may submit proposed changes for revi…".

    What it means: Under Agent access, an access profile sets the agent's baseline. Standard is selected. It reads authorized knowledge and can submit proposed changes for review, but has no approval authority.

  6. Click RestrictedTo give this agent less, choose Restricted.
    1. Click "Restricted". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001).

    What it means: To give this agent less, choose Restricted.

  7. Click Set access profileThe change isn't applied until you save it. Select Set access profile.
    1. Click "Set access profile". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001).

    What it means: The change isn't applied until you save it. Select Set access profile.

  8. Review now restrictedThe profile is now Restricted, which lists three effective capabilities.
    1. Review "Now Restricted": "Restricted Full ordinary lifecycle control over its own task memory; read authorized info…".

    What it means: The profile is now Restricted, which lists three effective capabilities.

  9. Review agent governanceThese settings control what this agent can read and submit through DM. They don't control what an external agent does elsewhere, and approving durable knowledge always stays with a person.
    1. Review "Agent governance": "Agent governance New Agents start with full governed operational access. Customize access…".

    What it means: These settings control what this agent can read and submit through DM. They don't control what an external agent does elsewhere, and approving durable knowledge always stays with a person.

Only the access profile is changed. Individual capabilities, custom grants, credentials, connection setup and the readiness test are not demonstrated (see the agent connection episodes).

That an external agent honors these settings outside DM is not claimed or verified. The agent here is invented and was never connected to a real client.

Descriptive transcript

Read the full transcript, including what happens on screen

Demo environment · synthetic workspace and data · backend responses are simulated. See and narrow what an agent can reach
This guide uses a synthetic workspace with one invented agent and simulated backend responses. Changing the access profile here only changes the simulated record, and the on-frame banner says so; it proves nothing about a deployed backend. DM controls what the agent can read and submit through DM; it does not control what an external agent does elsewhere.

A banner reading "Demo environment · synthetic workspace and data · backend responses are simulated" stays on screen for the whole video.

00:00–00:06 · start

On screen: Start on DM Home (/). Click "Settings". The next view is "Workspace settings" (/settings/workspace). A highlight marks "Settings".

Narrator: Agents are managed from Settings.

Narrator: Select Settings in the sidebar.

00:06–00:09 · open-agents

On screen: Click "Agents". The next view is "Agents" (/control-panel/agents). A highlight marks "Agents".

Narrator: Select Agents.

00:09–00:16 · select-agent

On screen: Click "Select agent (Claude Code)". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001). A highlight labelled "Open the agent" marks "Open".

Narrator: This workspace has one agent, Claude Code.

Narrator: Open it to see what it is allowed to do.

00:16–00:22 · summary

On screen: A highlight labelled "Capabilities enabled" marks "Access 12 capabilities enabled Governed capabilities this Agent may u…".

Narrator: The Access step summarizes how many capabilities this agent has enabled.

00:22–00:38 · profiles

On screen: A highlight labelled "Current profile" marks "Standard Read access to authorized shared knowledge; may submit propo…".

Narrator: Under Agent access, an access profile sets the agent's baseline.

Narrator: Standard is selected. It reads authorized knowledge and can submit proposed changes for review, but has no approval authority.

00:38–00:43 · choose

On screen: Click "Restricted". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001). A highlight labelled "Restricted" marks "Restricted Full ordinary lifecycle control over its own task memory;…".

Narrator: To give this agent less, choose Restricted.

00:43–00:49 · save

On screen: Click "Set access profile". The next view is "Claude Code" (/control-panel/agents/agent-claude-code-001). A highlight marks "Set access profile".

Narrator: The change isn't applied until you save it.

Narrator: Select Set access profile.

00:49–00:56 · result

On screen: A highlight labelled "Now Restricted" marks "Restricted Full ordinary lifecycle control over its own task memory;…".

Narrator: The profile is now Restricted, which lists three effective capabilities.

00:56–01:09 · boundary

On screen: A highlight labelled "Agent governance" marks "Agent governance New Agents start with full governed operational acce…".

Narrator: These settings control what this agent can read and submit through DM.

Narrator: They don't control what an external agent does elsewhere, and approving durable knowledge always stays with a person.

Last verified against the product on .