Docs / Agents

Agent status

How Tezgah knows what each agent is doing, which agents it supports, and how subagents are shown.

Every session in the sidebar has a status dot and a line saying what its agent is doing.

Status Dot Means
Working blue The agent is busy: reading, editing, running a command.
Needs you orange It’s asking for permission or waiting for an answer.
Done green It finished and is waiting for your next message.
New output grey A shell printed something while you were elsewhere.

How each agent reports

No setup is needed for any of them.

  • Claude Code reports its own state to Tezgah (over OSC 7501), so its status is exact: which tool it’s running, on which file, and when it needs you.
  • Codex reports exactly through hooks Tezgah passes when you run codex in a Tezgah terminal. Codex asks you to trust them once; your Codex config isn’t changed.
  • Gemini, Grok, Aider, OpenCode, Amp and others are read from what they do: output while working, a question when they need you, quiet when done. This works on macOS and Linux.

Each pane is named after the agent working in it, so a shell where you started gemini shows as Gemini.

Set the status yourself

Any script can set its pane’s status with the tezgah command:

tezgah status working "deploying"
tezgah status done
tezgah status clear

Subagents

When Claude starts subagents, they’re listed under their session in the sidebar, each with its own status line. Click one to follow its transcript as it works.

Plan limits, context and tokens

The bottom of the sidebar shows your Claude and Codex plan limits, and each Claude session’s header shows how full its context is. See Plan limits & usage.

Something missing? Open an issue