ThreeRouter
User Guide

1. Open VS Code and select any project folder to enter the workspace.
2. Open the Extensions panel from the left sidebar.
3. Search for "Codex" and click Install.
4. The Codex icon will appear after installation.
5. Note: If you previously logged in with official or other platforms, please log out first before reconfiguring!

Using Official Terminal: Install Codex CLI

npm i -g @openai/codex

Configuration Without CC Switch (Windows)

Configure config.toml

Create (if not exists) or edit C:\Users\{username}\.codex\config.toml:

model_provider = "Deepseek"
model = "deepseek-v4"
review_model = "deepseek-v4"
model_reasoning_effort = "xhigh"
disable_response_storage = true
network_access = "enabled"
windows_wsl_setup_acknowledged = true
model_context_window = 1000000
model_auto_compact_token_limit = 900000

[model_providers.Deepseek]
name = "Deepseek"
base_url = "https://api.threerouter.com"
wire_api = "responses"
requires_openai_auth = true

Configure auth.json

Create (if not exists) or edit C:\Users\{username}\.codex\auth.json:

{
  "OPENAI_API_KEY": "Replace with your API Key"
}

The official extension currently uses Google Cloud (GCP) and requires logging into your Google account to select a project space. Therefore, the official extension is not recommended.

💻 Using Official Terminal: Install Gemini CLI

npm install -g @google/gemini-cli

Configuration Without CC Switch (Windows)

Configure .env

Create (if not exists) or edit C:\Users\{username}\.gemini\.env:

GEMINI_API_KEY=Replace with your API-KEY 【Do not include quotes or special characters】
GOOGLE_GEMINI_BASE_URL=https://api.threerouter.com

Configure settings.json

Create (if not exists) or edit C:\Users\{username}\.gemini\settings.json:

{
  "security": {
    "auth": {
      "selectedType": "gemini-api-key"
    }
  }
}

OpenCode is special and suitable for users without fixed model preferences. The official extension is not yet complete and not recommended.

Using Official Terminal: Install OpenCode CLI

npm i -g opencode-ai

Configuration Without CC Switch (Windows)

Configure opencode.json

Create (if not exists) or edit C:\Users\{username}\.config\opencode\opencode.json:

{
  "provider": {
    "gemini": {
      "options": {
        "baseURL": "https://api.threerouter.com/v1beta",
        "apiKey": "Replace with your API-KEY"
      },
      "npm": "@ai-sdk/google",
      "models": {
        "deepseek-v4": {
          "name": "DeepSeek-V4",
          "limit": {
            "context": 1048576,
            "output": 65536
          },
          "modalities": {
            "input": ["text", "image"],
            "output": ["text"]
          }
        }
      }
    }
  },
  "$schema": "https://opencode.ai/config.json"
}

Trae is an IDE that supports multiple AI models. Configure ThreeRouter using the following steps:

Add Custom Model

Add custom model configuration in Trae:

Model Configuration

Configure OpenAI-type model connection:

Test after configuration:

1. Open the API Keys page and copy your API-key. (Choose different groups according to your model needs)
2. Open OpenClaw config file: C:\Users\{username}\.openclaw\openclaw.json

"models": {
  "mode": "merge",
  "providers": {
    "threerouter": {
      "baseUrl": "https://api.threerouter.com/v1",
      "apiKey": "Replace with your API-KEY",
      "api": "openai-completions",
      "models": [
        {
          "id": "deepseek-v4",
          "name": "deepseek-v4",
          "contextWindow": 200000,
          "maxTokens": 8192
        }
      ]
    }
  }
}

Find the "models" key in the config file and replace its content with the above. Do not modify other configuration. After changing the config, restart with openclaw gateway restart in the terminal.

1. Open the API Keys page and copy your API-key. (Choose different groups according to your model needs)
2. Open Hermes config file: ~/.hermes/config.yaml

model:
  default: deepseek-v4
  provider: custom
  base_url: "https://api.threerouter.com/v1"
  api_key: "Enter your API-KEY"

Find the config file and replace the model section with the above content. Do not modify other parts. Pay attention to the format. Restart with hermes gateway restart after saving.

/compact [instructions]  # You can add instructions

# Resume recent conversation
claude --continue

# Show interactive conversation selector
claude --resume

> I need to implement a new authentication system using OAuth2 for our API. Deep think about the best way to implement this in our codebase.

> Think about potential security vulnerabilities in this approach

> Think deeper about edge cases we should handle

CC Switch Model Switching

CC Switch allows easy switching between different models. Here are two common methods:

Common Slash Commands

Image Generation

ThreeRouter supports image generation via API. Here are the steps:

Step 1: Open the test key page to confirm image generation is available:

Step 2: Send image generation request and view results:

Step 3: Use curl script to call image API:

Step 4: Use Python script to call image API:

Recharge

Recharge entrance features:

Referral Bonus

Invite friends to register and earn rewards:

Real vs Fake Official Relay Comparison

Below is a comparison of real and fake official relays to help you identify reliable relay services:

Real official relay error screenshot:

Fake official relay no error return:

Fake official relay continues conversation:

npm Installation

Some tools need to be installed via npm. Below is the npm installation download page:

Windows:

Download: https://nodejs.org/en/download/

Verify: Open a new terminal and enter npm --version. If a version number appears, installation is complete.

Macos / Linux:

# Download and install fnm:
curl -o- https://fnm.vercel.app/install | bash

# Download and install Node.js:
fnm install 24

# Verify the Node.js version:
node -v # Should print "v24.15.0".

# Verify npm version:
npm -v # Should print "11.12.1".

GPT-Image-2 Image Generation

ThreeRouter supports OpenAI GPT-Image-2 through the OpenAI provider. You can call it using the standard OpenAI /v1/images/generations endpoint.

Supported models:

• gpt-image-2 - text-to-image generation

Account Configuration

In the admin dashboard, add an OpenAI API Key account with these settings:

• Base URL: https://api.threerouter.com/v1
• API Key: your threerouter API key
• Model Whitelist: gpt-image-2

Text-to-Image Example

curl -X POST "https://api.threerouter.com/v1/images/generations" \
  -H "Authorization: Bearer <your-api-key>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "a cute cat astronaut on a pastel background"
  }'

Response Format

By default the response contains a URL. Some browsers may block direct access to the OpenAI OSS URL due to referer restrictions, so the URL is best used server-side.

{
  "created": 1234567890,
  "data": [
    {
      "url": "https://img.oss-us-west-1.aliyuncs.com/images/xxx.png"
    }
  ]
}

Test: curl -I -H "User-Agent: Mozilla/5.0" "https://img.oss-us-west-1.aliyuncs.com/images/zzz.png"

Returning Base64

To receive the image as a base64 string instead, set response_format to b64_json. This is useful for front-end clients that cannot access the OSS URL directly.

curl -X POST "https://api.threerouter.com/v1/images/generations" \
  -H "Authorization: Bearer <your-api-key>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "a cute cat astronaut on a pastel background",
    "response_format": "b64_json"
  }'

Notes

• Image generation is asynchronous and may take 30-90 seconds.
• Only the /v1/images/generations endpoint is supported for image generation.
• If the edit endpoint does not accept url, try response_format=b64_json with a base64 string instead.