All settings

This list is generated from the code, so it matches the installed version. Settings explains how to change them, which source wins, and when a change takes effect.

Settings

Set any of these with claudio set <key> <value>, in a project's .claudio.json, or with its environment variable:

orchestratortext

the session that gets tasks by default, by title (a cmux workspace or a bb thread)

default: ""
accepts: text, may be empty
env: CLAUDIO_ORCHESTRATOR

modeltext

the brain's model, through your Claude Code login

default: default
accepts: default · sonnet · opus · haiku
env: CLAUDIO_MODEL

efforttext

how hard the brain thinks before it answers

default: low
accepts: low · medium · high
env: CLAUDIO_EFFORT

voicetext

any Deepgram Aura-2 voice: a name such as zeus, or a full model name

default: aura-2-orion-en
accepts: text
env: CLAUDIO_VOICE

earstext

Deepgram Flux model; flux-general-multi for other languages

default: flux-general-en
accepts: text
env: CLAUDIO_EARS

eager_eotnumber

confidence at which the brain starts on a turn early, before you have surely finished

default: 0.3
accepts: 0 to 1
env: CLAUDIO_EAGER_EOT

eotnumber

confidence at which your turn counts as finished

default: 0.7
accepts: 0 to 1
env: CLAUDIO_EOT

eot_timeout_msinteger

silence after which your turn ends whatever the confidence

default: 1500
accepts: 100 to 10000
env: CLAUDIO_EOT_TIMEOUT_MS

barge_inboolean

talking over it pauses it

default: true
accepts: true · false
env: CLAUDIO_BARGE_IN

memoryboolean

resume the last conversation, up to a week old

default: true
accepts: true · false
env: CLAUDIO_MEMORY

push_to_talkboolean

the mic stays closed until you press Enter, for one turn

default: false
accepts: true · false
env: CLAUDIO_PUSH_TO_TALK

address_meboolean

only turns that start with its name count, for a room with other people

default: false
accepts: true · false
env: CLAUDIO_ADDRESS_ME

narratetext

which sessions' finishes to narrate

default: handed-off
accepts: handed-off · all · none
env: CLAUDIO_NARRATE

mouthtext

who speaks: Deepgram Aura, ElevenLabs, or local (Kokoro on this Mac, no key)

default: deepgram
accepts: deepgram · elevenlabs · local
env: CLAUDIO_MOUTH

styletext

how it talks: plain, brief (one sentence) or caveman

default: plain
accepts: plain · brief · caveman
env: CLAUDIO_STYLE

elevenlabs_voicetext

ElevenLabs voice id, when mouth is elevenlabs

default: pNInz6obpgDQGcFmaJgB
accepts: text
env: CLAUDIO_ELEVENLABS_VOICE

local_voicetext

Kokoro voice, when mouth is local: am_michael, af_heart, bm_george…

default: am_michael
accepts: text
env: CLAUDIO_LOCAL_VOICE

local_speednumber

how fast the local voice speaks, 0.5 to 2.0

default: 1
accepts: 0.5 to 2
env: CLAUDIO_LOCAL_SPEED

local_modeltext

Kokoro model: compressed (88 MB, about twice as fast) or full (325 MB, a little cleaner)

default: compressed
accepts: compressed · full
env: CLAUDIO_LOCAL_MODEL

code_colorsboolean

highlight code blocks in colour; off shows them as plain text

default: true
accepts: true · false
env: CLAUDIO_CODE_COLORS

layouttext

inline, or split: your turns on the right, claudio's on the left

default: inline
accepts: inline · split
env: CLAUDIO_LAYOUT

nametext

what it answers to, and calls itself: one to three words

default: claudio
accepts: text
env: CLAUDIO_NAME

backendtext

where the sessions are: cmux, or bb (its threads)

default: cmux
accepts: cmux · bb
env: CLAUDIO_BACKEND

backend picks where your sessions live, and with it what orchestrator names: a cmux workspace's title or a bb thread's title. Platforms compares them.

The keys live in secrets.json, readable only by you. A key put in config.json is moved to secrets.json, and one in a project's .claudio.json is ignored. The brain may not read secrets.json, and the settings it is shown leave the keys out.

claudio setup stores the Deepgram key. Add the ElevenLabs key to secrets.json yourself, or use the environment variables below:

deepgram_api_keytext

Deepgram key, for ears and the Aura mouth

accepts: text
env: DEEPGRAM_API_KEY

elevenlabs_api_keytext

ElevenLabs key, when mouth is elevenlabs

accepts: text
env: ELEVENLABS_API_KEY

Settings

API keys

On this page

All settings

SettingsCopy Anchor Link

API keysCopy Anchor Link

On this page

Settings

API keys