Antigravity CLI + LLMoxy
Google's terminal agent and the successor to Gemini CLI. In API key mode, model requests go straight to LLMoxy.
Quick Start
- 1
Install Antigravity CLI
Run
agy --versionto confirm the install. If the command is not found, add~/.local/binto your PATH.BASH - 2
Create an API key
Create a token in the Console, pick the group you want to use, and copy the key.
Create a token - 3
Switch to API key mode
Add this to
~/.gemini/antigravity-cli/settings.json. If the file already exists, merge this entry into it. Without it, the environment variables alone have no effect.JSON - 4
Set the URL and key
The CLI reads the key only from environment variables and does not load
.envfiles. The Gemini-compatible URL has no/v1. To keep these settings, add both lines to~/.zshrcor~/.bashrc.BASH - 5
Get started
In the same terminal, run
agyin your project directory. When the header shows Gemini API key, you are connected. Type/modelto switch models.BASH
Tips
- If you used Gemini CLI before, the first
agyrun offers to import your old configuration, and your project'sGEMINI.mdstill applies. - To go back to signing in with a Google account, remove
modelProviderfromsettings.jsonand restart the CLI.
FAQ
The CLI exits at startup saying GEMINI_API_KEY is not set?
modelProvider is set to gemini, but the current terminal does not have this environment variable. Export it again, or add it to your shell profile and open a new terminal.Starts fine but errors during a conversation?
Other tools
yetone
magpie
Pick models for all your agents in one place. Claude Code, Codex, OpenCode, and others each use their own model, routed through one local gateway.
Anthropic
Claude
Anthropic's AI assistant and coding agent: Claude Desktop, plus Claude Code in the terminal and in VS Code.
OpenAI
ChatGPT (Codex)
OpenAI's coding agent. The ChatGPT desktop app, the CLI, and the IDE extension share one config.toml, so one setup covers all three.
Anomaly
OpenCode
Open-source coding agent with a desktop app and a terminal UI. You can configure multiple model providers.
DeepSeek
DeepSeek Harness
DeepSeek's open-source agent framework, with a desktop app, a web UI, and the dsh command line. It connects to OpenAI- or Anthropic-compatible model services.
Anysphere
Cursor
AI editor built on VS Code. Override the OpenAI URL to route OpenAI models in chat through LLMoxy.
View guideGitHub
GitHub Copilot
The Copilot app, Copilot CLI, and VS Code all support bring your own key (BYOK), with no Copilot subscription required.
ByteDance
Trae
ByteDance's AI IDE. Both the China and international versions support custom models.
View guideCline
Cline
Open-source coding agent with a full agent loop, available as a command-line tool and as VS Code and JetBrains extensions.
Alibaba
Qoder
Alibaba's AI coding product. The Qoder desktop app and Qoder CLI support custom models; Qoder IDE doesn't yet.
Tencent
CodeBuddy
Tencent's AI coding tools. Both the command-line CodeBuddy Code and CodeBuddy IDE support OpenAI-compatible custom models.
Tencent
WorkBuddy
Tencent's desktop AI workspace for office and coding tasks. It supports custom model services.
View guideCherryHQ
Cherry Studio
Cross-platform desktop AI client. Add a provider to fetch the list of available models in one click.
View guideChatbox AI
Chatbox
AI chat client for desktop and mobile. Add a model provider in settings to connect.
View guideLangGenius
Dify
Open-source LLM app development platform. Connect it as OpenAI-API-compatible to build apps and workflows.
View guiden8n
n8n
Workflow automation platform. Enter the LLMoxy URL in an OpenAI credential to call models.
View guide