Agentbeamdocs

Start here

MDM / Getting startedDevice managementOne place to understand your Apple and Windows fleet, deliver configuration, and request device actions.MDM / Getting startedSet up your workspaceRequest MDM access, check platform readiness, and get your first managed device into inventory.Agent CLI / Getting startedAgent CLI overviewUnderstand agent activity locally, scan the files your assistants use, and connect shared workspace controls when needed.Agent CLI / Getting startedCLI quickstartInstall Beam, capture a first event, and open your local activity dashboard.Agent CLI / Getting startedConnect a workspacePair your machine through the browser or an enrollment code, then verify identity and policy synchronization.
↑ ↓ to navigate · Enter to openEsc to close
Website Open dashboard
Agent CLI/Capture & investigate
Capture & investigate

Local Studio

Search captured activity, review findings, and scan content from the collector’s built-in browser interface.

On this page

Studio is served by the collector itself at / and /studio. It is a local interface with no separate installation or CDN requirement. Start the collector before opening it:

Code
beam start
# In another terminal
beam studio

The CLI opens your default browser at the configured loopback origin. If you changed ports, point both commands at the same collector configuration.

Pair your browser

Studio's page shell can load without authentication, but API access needs the collector pairing token. beam studio reads the token from disk and passes it through the opening URL. Studio reads it, removes it from the address bar, and stores it in that origin's browser local storage.

Do not share the initial URL. Use Disconnect to clear the browser's saved pairing. A rejected token is cleared so you can pair again. Different ports are different browser origins and need their own pairing state.

Investigate activity

The Activity view lists captured events with search and agent/risk filters. Open an event's detail drawer to inspect the normalized action, findings, and relevant evidence. Read event phase carefully: a proposed tool call is different from an action observed in a transcript.

To confirm end-to-end capture, perform a harmless tool action in a supported assistant and locate its record. An empty view may mean there are no retained events, the hook is not firing, the collector is unreachable, or pairing has failed.

Review findings

The Findings view focuses on events with heuristic matches. Mark an event reviewed after assessing its evidence. That marker records your review; it does not approve the command, reverse an effect, or change policy.

Connected installations can forward review changes to the workspace on a best-effort basis. Check both surfaces when a remote record appears stale.

Scan skills and MCP content

Use Skill & MCP scan to inspect pasted content or file contents with the same scanning engine as beam scan. Select the correct kind so MCP configuration receives JSON and version-pinning checks. Treat findings as review leads, not a malware verdict.

Current limits

Studio's core views are Activity, Findings, and Scan. Session/usage views from earlier prototypes are not part of this implementation. Local storage is bounded to 10,000 events and 500 scan reports; export evidence before retention rolls it out if you need to preserve it.

Keep exploring

Explore the docs