Skip to main content
This page is for admins once Bridge is rolled out: where to see what governed sessions do, and what to do when someone reports a problem. For what people see on their side, and what they can fix themselves, see What’s blocked and why.

Where Bridge traffic shows up

Every governed session sends its model traffic through your LLM Gateway and its tool calls through Barndoor, so Bridge usage from every client appears in the same places as the rest of your organization’s traffic. Bridge tags each session with its client settings version, so a change you save to client settings shows up as a new row from the next launch, or from a running session’s next hourly renewal. The AI Agent each client’s traffic is recorded under: Claude Desktop launched from Bridge shares the Claude Code agent and client settings with Claude Code, so the two can’t be told apart in reporting. Your policies can target these agents like any other. With Bridge governing MCP only, model traffic doesn’t go through the LLM Gateway, so it isn’t on the LLM Usage Dashboard. With Models only, there are no Barndoor tool calls. What doesn’t show up: what a client’s own built-in tools do on the laptop, such as a native shell command or file edit, except where it appears in the conversation sent to the model. To route those through Barndoor, see Client settings.

Launches that failed

LLM Management → Client settings lists, at the bottom, launches that failed in the last 7 days, grouped by client. Each row has the time, the user, the reason, how often it repeated, and the message the person saw. It’s the first place to look when someone says Bridge won’t launch.
Recent failed launches: two Claude rows with the reason Model access denied and the message None of the models you're allowed to use fit Claude Code, and one Codex row with the reason CLI out of date

Diagnose a laptop

Ask the person for two things:
  1. Bridge’s log. In the Bridge window, Logs → Copy. On a Mac it’s also at ~/Library/Logs/Barndoor/gui.log; on Windows, in %LOCALAPPDATA%\Barndoor\Logs.
  2. barndoor doctor in a terminal. It shows the Bridge version, the Barndoor environment, which managed settings arrived from your MDM, whether direct access to provider APIs is blocked, whether the barndoor command is Bridge’s, and the last lines of Bridge’s log.
The top of barndoor doctor output: the Bridge version, configuration with product, environment, API URL and launch settings, managed preferences set by MDM including publisher_keys, the tenant publisher key with its fingerprint and allowed hosts, and the direct inference egress check
To check whether a tool is allowed for that person, run barndoor tools test-policy --harness claude <server> <tool> on their laptop (or --harness codex), for example barndoor tools test-policy --harness claude filesystem read_text_file. Without --harness, the check ignores policies granted to a specific agent. It prints LIVE: ALLOWED or LIVE: DENIED.

Common problems