Onyx Terminal Update

hey. big drop today. onyx is now a real terminal app.

you can just install It and run instead of trying to put this into vs code.

WHAT IS NEW:

• one-command install on windows / mac / linux

• first-run wizard walks you through your API key

no config files

• four provider slots now:

─ anthropic (claude direct)

─ openrouter (one key, everything)

─ openai (gpt-5.6 sol / luna / terra,

gpt-4o, o1)

─ custom (deepseek, xai/grok, groq,

mistral, ollama, azure...

any OpenAI-compatible

endpoint — bring your key)

• gpt-5.6 family at the top of /model picker

• /jailbreak now targets gpt-5.6-sol-pro out of the

box

• your provider choice persists across launches

• state lives in ~/.onyx/ — safe to move, back up,

or wipe

GET IT RUNNING (pick your platform)

▸ WINDOWS

download → onyx.exe

double-click.

→ SmartScreen warning first time = normal

(unsigned build).

click "more info" → "run anyway". one-time only.

▸ MAC

download → onyx_tui-0.2.0-py3-none-any.whl

open Terminal, then:

pip3 install

~/Downloads/onyx_tui-0.2.0-py3-none-any.whl

onyx

(no Python? run: brew install python first)

▸ LINUX

download → onyx_tui-0.2.0-py3-none-any.whl

open a terminal, then:

pip install ./onyx_tui-0.2.0-py3-none-any.whl

onyx

FIRST LAUNCH (takes around 30 seconds)

onyx asks:

1) anthropic claude direct

2) openrouter one key, all models

3) openai gpt-5.6 / gpt-4o direct

4) custom any OpenAI-compatible endpoint

pick a number, paste your key, hit enter.

that's it, key is saved forever. next launch just opens the chat.

LIVE INSIDE ONYX (commands features etc..)

/help everything

/model switch model on the fly

/provider <name> switch backend

/key set <prov> <key> add another provider

/agent on tool-use mode (bash,

files, python, web) (THIS HAS NOT BE TESTED AS MUCH)

/jailbreak <objective> to jailbreak models with your objective

/save <name> persist convo

/load <name> restore convo

/quit to exit