Use the company Gemini 4 Argon in your coding agent

Gemini 4 Argon runs on Google Vertex AI in the company's Google Cloud project. You can use it from Claude Code, Codex, Hermes or any tool that speaks the OpenAI or Anthropic API. You need a personal API key; you get one by signing in with your Lark account. No tickets, no waiting.

1. Quick setup: one command (macOS, Linux, WSL)

Open a terminal and run:

curl -fsSL https://gemini-dev.7.solutions/setup.sh | bash

The script shows a short code and opens your browser.
Sign in with Lark, check that the code matches, and click Approve.
The script receives a new personal key and sets up the agents it finds on your computer (Claude Code, Codex, Hermes). Your existing settings are not changed.

Then start your agent:

Claude Code	`claude-gemini` (your normal `claude` stays as it is)
Codex	`codex-gemini`, or `codex --profile gemini7`
Hermes	`hermes chat --provider gemini7`, or pick “Gemini 4 Argon (7 Solutions)” in `hermes model`

Script options

# only some agents
curl -fsSL https://gemini-dev.7.solutions/setup.sh | bash -s -- --claude --codex
# also make Gemini 4 Argon the default model in Hermes
curl -fsSL https://gemini-dev.7.solutions/setup.sh | bash -s -- --hermes --hermes-default
# remove everything the script added (then revoke the key on the My keys page)
curl -fsSL https://gemini-dev.7.solutions/setup.sh | bash -s -- --uninstall

The script is plain bash: read it first with curl -fsSL https://gemini-dev.7.solutions/setup.sh | less. It stores the key in ~/.config/gemini7/credentials (readable only by you) and creates the wrappers in ~/.local/bin.

2. Manual setup (Windows, or if you prefer)

Get a key Sign in with Lark on the My keys page and create a key. It is shown once; copy it. In the snippets below, replace YOUR_KEY with it. After you create a key, the page shows the same snippets with the key already filled in.

Claude Code

Claude Code talks to the gateway's Anthropic-compatible endpoint. Put this in your shell profile (~/.zshrc or ~/.bashrc), or in a small wrapper script so your normal Claude login stays untouched:

export ANTHROPIC_BASE_URL=https://gemini-api.7.solutions/anthropic
export ANTHROPIC_AUTH_TOKEN=YOUR_KEY
export ANTHROPIC_MODEL=vertex/gemini-4-argon
export ANTHROPIC_DEFAULT_OPUS_MODEL=vertex/gemini-4-argon
export ANTHROPIC_DEFAULT_SONNET_MODEL=vertex/gemini-4-argon
export ANTHROPIC_DEFAULT_HAIKU_MODEL=vertex/gemini-4-argon
export CLAUDE_CODE_SUBAGENT_MODEL=vertex/gemini-4-argon
export CLAUDE_CODE_MAX_CONTEXT_TOKENS=262144
export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1
export API_TIMEOUT_MS=600000
claude

Every model slot (Opus, Sonnet, Haiku, sub-agents) points at Gemini 4 Argon, so nothing is sent to Anthropic. CLAUDE_CODE_MAX_CONTEXT_TOKENS sets when Claude Code compacts; it can go up to 1048576, but every token is billed.

Codex CLI

Create the profile file ~/.codex/gemini7.config.toml (Codex 0.134 or newer; your main config.toml is not touched):

model_provider = "gemini7"
model = "vertex/gemini-4-argon"
model_context_window = 262144

[model_providers.gemini7]
name = "Gemini 4 Argon (7 Solutions)"
base_url = "https://gemini-api.7.solutions/v1"
env_key = "GEMINI7_API_KEY"
wire_api = "responses"

Then run Codex with that profile:

export GEMINI7_API_KEY=YOUR_KEY
codex --profile gemini7

Codex may print “Model metadata … not found, defaulting to fallback metadata”. That warning is expected for a gateway model id and harmless.

Hermes Agent

Adds a provider named gemini7. The key goes to Hermes' .env, never to config.yaml:

hermes config set providers.gemini7.name "Gemini 4 Argon (7 Solutions)"
hermes config set providers.gemini7.api https://gemini-api.7.solutions/v1
hermes config set providers.gemini7.key_env GEMINI7_API_KEY
hermes config set providers.gemini7.transport chat_completions
hermes config set providers.gemini7.default_model vertex/gemini-4-argon
hermes config set providers.gemini7.context_length 262144
hermes config set GEMINI7_API_KEY YOUR_KEY

Use it once with hermes chat --provider gemini7 -m vertex/gemini-4-argon, or make it your default with hermes model and pick “Gemini 4 Argon (7 Solutions)”.

Any other tool (OpenAI or Anthropic compatible)

OpenAI base URL	`https://gemini-api.7.solutions/v1` (Chat Completions and Responses)
Anthropic base URL	`https://gemini-api.7.solutions/anthropic` (Messages)
Model	`vertex/gemini-4-argon`
API key	your personal key, as `Authorization: Bearer` or `x-api-key`

curl https://gemini-api.7.solutions/v1/chat/completions \
  -H "Authorization: Bearer YOUR_KEY" -H "Content-Type: application/json" \
  -d '{"model":"vertex/gemini-4-argon","messages":[{"role":"user","content":"Hello"}]}'

Windows PowerShell: set the same variables with $env:NAME = "value" (for example $env:ANTHROPIC_AUTH_TOKEN = "YOUR_KEY"), or permanently with setx NAME value. The Codex profile file goes to %USERPROFILE%\.codex\gemini7.config.toml.

3. Good to know

The company pays per token. Gemini 4 Argon costs $2 per million input tokens and $10 per million output tokens until 30 Nov 2026, then $4 and $20. A long agent session can use millions of tokens, so please use it for real work, not for unattended loops or bulk jobs.
Context. Up to 1M tokens (input and output share the window). The snippets set 262144 so agents compact earlier, which keeps cost down; raise it when you really need more.
Availability. Served by Google (Vertex AI, global endpoint); no company GPU involved.
Your keys. Up to 5 keys per person (one per computer is a good habit). Each key allows 120 requests per minute and expires after 90 days; run the setup again to get a fresh one. Revoke a lost key right away on My keys.
Privacy. Requests go from the company Google Cloud project to Vertex AI under Google Cloud's terms. The gateway does not store prompts or answers.
Browser chat. For plain chat, use https://gemini.7.solutions (Lark login, no key needed).

4. Troubleshooting

HTTP 401 “virtual key …”	The key is wrong, revoked or expired. Run the setup command again, or create a new key.
HTTP 429	More than 120 requests in a minute with one key, or Vertex AI is busy. Slow down or wait a minute.
HTTP 5xx	A temporary Vertex AI error. Try again; the gateway already retries twice.
Hermes: 401 right after setup	Restart the Hermes gateway so it re-reads `.env`: `hermes gateway restart`.
`claude-gemini: command not found`	Add `~/.local/bin` to your PATH: `echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.zshrc`, then open a new terminal.