- Docs
- Agents and consoles
- Live agent status
Live agent status
A terminal alone cannot tell an agent that is thinking from one that is waiting for you. This setting lets Claude Code and Codex tell Symvibe themselves, through their own hooks.
On this page
Turn it on#
- Open the setting
Go to SettingsAgentsLive agent status and turn on Report status from the agent itself. It is off until you do, because it adds an entry to each CLI's own settings.
- Trust the hooks in Codex
Codex asks once to review the new hooks. Choose Trust all and continue.
- Restart running consoles
Consoles that were already running need a Restart from their header. New consoles pick it up on their own.
What it changes on your Mac#
- A Symvibe entry in
~/.claude/settings.jsonand~/.codex/hooks.json, or in the folders yourCLAUDE_CONFIG_DIRandCODEX_HOMEpoint to, and in each saved account profile. - A small script,
~/.hypervibe/agent-hooks/hypervibe-hook.sh, that passes each event to Symvibe over a local socket. The events stay on your Mac. - A one-time backup of each file it edits, taken before the first change:
settings.json.hypervibe-backupfor Claude Code andhooks.json.hypervibe-backupfor Codex, in your own folders and in each account profile. Your status line is never touched, and a settings file Symvibe cannot read is left alone. - Turning the setting off removes the entries.
What you see#
| The agent | Status pill |
|---|---|
| Starts a turn or a tool | WORKING |
| Asks for a permission | NEEDS YOU |
| Asks you a question | NEEDS YOU |
| Starts up, or finishes its turn | READY |
| Fails its turn | ERROR |
The question and failed-turn rows come from Claude Code; Codex's hooks report neither.
- Hover a pill to read what the agent says it is doing: the command it runs, the permission it asks for, the question it waits on.
- The needs you chip at the top right, and N, count and jump to consoles that need you, errors included.
- The Mission Map filters use the same states, and the loop trigger When an agent goes idle fires on them.
- Reports in the command bar's thread use the agent's final reply, up to 4,000 characters.
Without it#
WORKING then only means output arrived recently, so an agent showing a permission prompt with a spinner reads WORKING while it waits for you, and NEEDS YOU and READY never appear. Orders still wait while a permission or question is on screen, because Symvibe recognizes those prompts on the screen itself.
Which consoles#
Built-in Claude Code and Codex consoles. Gemini, Grok, OpenCode, catalog agents and your own commands keep the output-based status.
Keep going#
Something on this page does not match what you see in the app? Write to hello@symvibe.com or tell us in the Discord, and we will fix it.