Your agents run in a dozen terminals and you find out what they did afterwards, if at all.
agentglass puts every one of them on one screen: live tool calls, tokens and cost, across Claude Code, Codex, Gemini CLI, Kimi and anything that speaks OpenTelemetry.
It is a workspace, not just a dashboard. Review the diff, drive git, answer a chat, and hold a dangerous command until you approve it. From the desk or from your phone.
Runs on localhost. No account, no telemetry, never phones home. MIT.
