AI Coding Tools¶

Argo Proxy v3 serves all major LLM API formats, which means it works out of the box with popular CLI tools and AI coding assistants. This guide shows how to configure each tool to use argo-proxy as its backend.

Prerequisites

All examples assume argo-proxy is running at http://localhost:44497. Replace with your actual host and port.

Your ANL username is used as the API key for all tools.

Claude Code¶

Claude Code (tested with v2.1.81) connects via the Anthropic Messages API (/v1/messages).

Config File (Recommended)Environment Variables

claude config set -g apiKeyHelper "echo 'your-anl-username'"
claude config set -g env.ANTHROPIC_BASE_URL "http://localhost:44497"
claude config set -g env.CLAUDE_CODE_SKIP_ANTHROPIC_AUTH "1"

This writes to ~/.claude/settings.json:

{
    "apiKeyHelper": "echo 'your-anl-username'",
    "env": {
        "ANTHROPIC_BASE_URL": "http://localhost:44497",
        "CLAUDE_CODE_SKIP_ANTHROPIC_AUTH": "1"
    }
}

export ANTHROPIC_BASE_URL="http://localhost:44497"
export ANTHROPIC_API_KEY="your-anl-username"
export CLAUDE_CODE_SKIP_ANTHROPIC_AUTH=1
claude

Important

CLAUDE_CODE_SKIP_ANTHROPIC_AUTH=1 is required — it skips Anthropic's default authentication flow
Set ANTHROPIC_BASE_URL to the proxy root (e.g., http://localhost:44497), not http://localhost:44497/v1/messages — Claude Code appends the path automatically

Codex CLI (OpenAI)¶

Codex CLI (tested with v0.116.0) connects via the OpenAI Responses API (/v1/responses).

Config File (Recommended)Environment Variables

Add a custom provider in ~/.codex/config.toml:

model = "argo:gpt-5"
model_provider = "argo"

[model_providers.argo]
name = "Argo Proxy"
base_url = "http://localhost:44497/v1"
env_key = "ARGO_API_KEY"
wire_api = "responses"

Then set the API key in your shell profile:

export ARGO_API_KEY="your-anl-username"

export OPENAI_BASE_URL="http://localhost:44497/v1"
export OPENAI_API_KEY="your-anl-username"
codex

Note

Codex CLI uses the Responses API wire format by default. When using the config file approach, wire_api = "responses" makes this explicit.

Aider¶

Aider supports both OpenAI and Anthropic backends.

OpenAI ModeAnthropic Mode

export OPENAI_API_BASE="http://localhost:44497/v1"
export OPENAI_API_KEY="your-anl-username"
aider --model argo:gpt-5

export ANTHROPIC_BASE_URL="http://localhost:44497"
export ANTHROPIC_API_KEY="your-anl-username"
aider --model argo:claude-4.5-sonnet

Tip

You can add these to your .aider.conf.yml or shell profile for persistence.

Gemini CLI (Discontinued)¶

Discontinued

Gemini CLI stopped serving on 2026-06-18 and has been replaced by Google Antigravity CLI (agy). The section below is kept for reference only.

Gemini CLI connected via the Google GenAI API (/v1beta/models/{model}:generateContent).

Config FilesEnvironment Variables

1. ~/.gemini/.env:

GEMINI_API_KEY=your-anl-username
GOOGLE_GEMINI_BASE_URL=http://localhost:44497

2. ~/.gemini/settings.json:

{
    "model": {
        "name": "argo:gpt-4.1"
    },
    "security": {
        "auth": {
            "selectedType": "gemini-api-key"
        }
    }
}

GOOGLE_GEMINI_BASE_URL=http://localhost:44497 \
GEMINI_API_KEY=your-anl-username \
gemini -m argo:gpt-4.1

Google Antigravity CLI (agy)¶

Google Antigravity CLI (agy, tested with v1.1.20) is the successor to Gemini CLI. It connects via the Google GenAI API (/v1beta/models/{model}:generateContent), the same protocol as the former Gemini CLI.

Config Files (Recommended)Environment Variables

1. ~/.gemini/antigravity-cli/settings.json:

{
    "modelProvider": "gemini",
    "model": {
        "name": "argo:gpt-5"
    }
}

2. Environment variables (add to shell profile):

export GEMINI_API_KEY="your-anl-username"
export GOOGLE_GEMINI_BASE_URL="http://localhost:44497"

Then just run agy.

GOOGLE_GEMINI_BASE_URL=http://localhost:44497 \
GEMINI_API_KEY=your-anl-username \
agy

Planner model requirement

agy uses a hardcoded internal planner model (gemini-3.1-pro-preview) alongside your chosen model. This cannot be overridden via settings.json or CLI flags — there is no user-facing configuration for it. The model must be available on the upstream — argo-proxy maps it automatically if the upstream provides it.

Note

Config path changed from Gemini CLI's ~/.gemini/settings.json to ~/.gemini/antigravity-cli/settings.json
modelProvider must be set to "gemini" in the settings file
agy sends x-goog-api-key header for auth, which argo-proxy supports natively

OpenCode¶

OpenCode (tested with v1.2.27) supports OpenAI-compatible endpoints.

Config File (Recommended)Environment Variables

Add a custom provider in ~/.config/opencode/opencode.json:

{
    "provider": {
        "argo-proxy": {
            "npm": "@ai-sdk/openai-compatible",
            "name": "Argo Proxy",
            "options": {
                "baseURL": "http://localhost:44497/v1",
                "apiKey": "your-anl-username"
            },
            "models": {
                "argo:gpt-5": {
                    "name": "GPT-5"
                },
                "argo:claude-4.5-sonnet": {
                    "name": "Claude 4.5 Sonnet"
                }
            }
        }
    }
}

export OPENAI_BASE_URL="http://localhost:44497/v1"
export OPENAI_API_KEY="your-anl-username"
opencode

Kilo Code¶

Kilo Code (VS Code extension) supports Anthropic API.

Configure in Kilo Code settings:

Base URL: http://localhost:44497
API Key: your-anl-username

Generic OpenAI SDK¶

Any tool or script using the OpenAI Python SDK:

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:44497/v1",
    api_key="your-anl-username",
)

response = client.chat.completions.create(
    model="argo:gpt-5",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)

Generic Anthropic SDK¶

Any tool or script using the Anthropic Python SDK:

import anthropic

client = anthropic.Anthropic(
    base_url="http://localhost:44497",
    api_key="your-anl-username",
)

message = client.messages.create(
    model="argo:claude-4.5-sonnet",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}],
)
print(message.content[0].text)

Summary¶

Tool	API Format	Base URL Env Var	Value
Claude Code	Anthropic	`ANTHROPIC_BASE_URL`	`http://localhost:44497`
Codex CLI	OpenAI Responses	`OPENAI_BASE_URL`	`http://localhost:44497/v1`
Aider (OpenAI)	OpenAI	`OPENAI_API_BASE`	`http://localhost:44497/v1`
Aider (Anthropic)	Anthropic	`ANTHROPIC_BASE_URL`	`http://localhost:44497`
Gemini CLI (discontinued)	Google GenAI	`GOOGLE_GEMINI_BASE_URL`	`http://localhost:44497`
agy	Google GenAI	`GOOGLE_GEMINI_BASE_URL`	`http://localhost:44497`
OpenCode	OpenAI	`OPENAI_BASE_URL`	`http://localhost:44497/v1`
OpenAI SDK	OpenAI	`OPENAI_BASE_URL`	`http://localhost:44497/v1`
Anthropic SDK	Anthropic	`ANTHROPIC_BASE_URL`	`http://localhost:44497`

Note

For all tools, use your ANL username as the API key. The actual authentication is handled by the ARGO backend based on your network identity.