Skip to main content

​
Overview

Cline (formerly Claude Dev) is a VS Code extension that provides autonomous AI coding assistance. Cline supports OpenAI-compatible API endpoints, allowing you to use CLI Proxy API with your Google/ChatGPT/Claude OAuth subscriptions.

​
Configuration

1

Install Cline

Install the Cline extension from the VS Code marketplace:
  1. Open VS Code
  2. Go to Extensions (Cmd+Shift+X on macOS, Ctrl+Shift+X on Windows/Linux)
  3. Search for “Cline”
  4. Click Install
2

Start CLI Proxy API

Ensure CLI Proxy API is running:
./cliproxyapi
The server will listen on http://localhost:8317 by default.
3

Configure Cline API Settings

Open Cline settings in VS Code:
  1. Open Command Palette (Cmd+Shift+P on macOS, Ctrl+Shift+P on Windows/Linux)
  2. Type “Cline: Open Settings”
  3. Select OpenAI Compatible as the API provider
4

Set Endpoint and API Key

Configure the connection:
  • API Provider: OpenAI Compatible
  • Base URL: http://localhost:8317/v1
  • API Key: Use any key from your api-keys list in config.yaml
  • Model: Select from available models (e.g., gemini-2.5-pro, claude-sonnet-4)

​
Configuration Examples

​
Using Gemini OAuth

If you have Gemini CLI OAuth configured:
settings.json
{
  "cline.apiProvider": "openai-compatible",
  "cline.baseUrl": "http://localhost:8317/v1",
  "cline.apiKey": "your-api-key-1",
  "cline.model": "gemini-2.5-pro"
}

​
Using Claude OAuth

If you have Claude Code OAuth configured:
settings.json
{
  "cline.apiProvider": "openai-compatible",
  "cline.baseUrl": "http://localhost:8317/v1",
  "cline.apiKey": "your-api-key-1",
  "cline.model": "claude-sonnet-4"
}

​
Using OpenAI Codex

If you have OpenAI Codex OAuth configured:
settings.json
{
  "cline.apiProvider": "openai-compatible",
  "cline.baseUrl": "http://localhost:8317/v1",
  "cline.apiKey": "your-api-key-1",
  "cline.model": "gpt-5"
}

​
Advanced Configuration

​
Multiple API Keys

If you want to use different API keys for different projects:
config.yaml
api-keys:
  - "project-a-key"
  - "project-b-key"
  - "personal-key"
Then configure Cline per-project using workspace settings (.vscode/settings.json):
.vscode/settings.json
{
  "cline.apiKey": "project-a-key",
  "cline.model": "gemini-2.5-pro"
}

​
Model Prefixes

If you have multiple credentials with prefixes:
config.yaml
gemini-api-key:
  - api-key: "AIzaSy...01"
    prefix: "work"
  - api-key: "AIzaSy...02"
    prefix: "personal"
Use the prefix in your model selection:
settings.json
{
  "cline.model": "work/gemini-2.5-pro"
}

​
Custom Model Aliases

Create model aliases for easier switching:
config.yaml
oauth-model-alias:
  gemini-cli:
    - name: "gemini-2.5-pro"
      alias: "g2.5p"
  claude:
    - name: "claude-sonnet-4-5-20250929"
      alias: "cs4.5"
Then in Cline:
settings.json
{
  "cline.model": "g2.5p"  // Maps to gemini-2.5-pro
}

​
Features

​
Autonomous Coding

Cline can autonomously:
  • Read and analyze your codebase
  • Create, edit, and delete files
  • Run terminal commands
  • Search for information
  • Debug issues

​
Streaming Responses

Cline supports streaming responses through CLI Proxy API:
  • Real-time code generation
  • Progressive task execution
  • Instant feedback on actions

​
Function Calling

For models that support function calling (Gemini, OpenAI, Claude):
  • File system operations
  • Code analysis tools
  • Search and navigation
  • Terminal integration

​
Multimodal Support

For models that support images (Gemini, Claude):
  • Analyze screenshots
  • Debug UI issues
  • Design reviews

​
Workflow Examples

​
Example 1: Full-Stack Development

1

Configure Cline

settings.json
{
  "cline.baseUrl": "http://localhost:8317/v1",
  "cline.apiKey": "your-api-key-1",
  "cline.model": "gemini-2.5-pro"
}
2

Give Cline a Task

Open Cline panel and type:
Create a REST API with Express.js that has CRUD endpoints for a task management system
3

Review and Approve

Cline will:
  • Create necessary files
  • Write the code
  • Ask for approval before executing commands
4

Test and Iterate

Continue the conversation:
Add input validation and error handling

​
Example 2: Bug Fixing

1

Select Claude Model

settings.json
{
  "cline.model": "claude-sonnet-4"
}
2

Describe the Bug

The user authentication is failing with a 401 error. 
Check the auth middleware and fix the issue.
3

Let Cline Investigate

Cline will:
  • Read relevant files
  • Analyze the code
  • Propose fixes
  • Execute tests

​
Troubleshooting

​
Connection Errors

If Cline shows connection errors:
  1. Verify CLI Proxy API is running: 
    curl http://localhost:8317/v1/models
  2. Check the Base URL in Cline settings matches your config
  3. Ensure no firewall is blocking localhost connections

​
Authentication Failures

If you see “Invalid API key” or authentication errors:
  1. Verify the API key in Cline matches one in your config.yaml:
    config.yaml
    api-keys:
  - "your-api-key-1"
  2. Check for whitespace or special characters
  3. Restart CLI Proxy API after config changes

​
Model Not Available

If Cline can’t use a specific model:
  1. Authenticate with the provider first: 
    ./cliproxyapi gemini login
./cliproxyapi claude login
  2. Verify the model is listed: 
    curl -H "Authorization: Bearer your-api-key-1" \
  http://localhost:8317/v1/models | jq '.data[].id'
  3. Check provider configuration in config.yaml

​
Slow Responses

If responses are slower than expected:
  1. Check your network connection to OAuth providers
  2. Enable debug logging to see timing info:
    config.yaml
    debug: true
  3. Consider using faster models (e.g., gemini-2.5-flash instead of gemini-2.5-pro)
  4. Configure multiple accounts for load balancing

​
Best Practices

​
1. Choose the Right Model

  • Quick tasks: gemini-2.5-flash, claude-haiku-4
  • Complex reasoning: gemini-2.5-pro, claude-sonnet-4
  • Maximum capability: claude-opus-4, gpt-5

​
2. Use Workspace Settings

Create project-specific settings in .vscode/settings.json:
.vscode/settings.json
{
  "cline.apiKey": "project-specific-key",
  "cline.model": "gemini-2.5-pro"
}

​
3. Configure Multiple Providers

Authenticate with multiple providers for redundancy: 
./cliproxyapi gemini login
./cliproxyapi claude login
./cliproxyapi codex login

​
4. Enable Load Balancing

config.yaml
routing:
  strategy: "round-robin"
  
max-retry-credentials: 3

​
5. Monitor Usage

config.yaml
usage-statistics-enabled: true
Then check usage via Management API or logs.

​
Performance Optimization

​
Reduce Latency

config.yaml
# Enable streaming for faster first-token response
streaming:
  keepalive-seconds: 15
  bootstrap-retries: 1

​
Handle Rate Limits

config.yaml
quota-exceeded:
  switch-project: true
  switch-preview-model: true
  
max-retry-credentials: 3
max-retry-interval: 30

​
Optimize for Concurrent Requests

config.yaml
# Enable commercial mode for high-concurrency environments
commercial-mode: true

# Configure multiple accounts
gemini-api-key:
  - api-key: "key1"
  - api-key: "key2"
  - api-key: "key3"

​
See Also

Build docs developers (and LLMs) love

Get started for freeTalk to us