Continue.dev is an open-source AI coding assistant extension for VS Code and JetBrains IDEs. Unlike GitHub Copilot (Microsoft-only model) or Cursor (proprietary IDE), Continue is model-agnostic: use Claude, GPT-4o, Gemini, or completely local models via Ollama — with no data sent anywhere if using local models. It’s the most flexible AI coding tool for developers who want control over their AI infrastructure.

Why Continue Over GitHub Copilot?

Feature	GitHub Copilot	Continue.dev
Model choice	GitHub/OpenAI only	Any LLM (Claude, GPT, Gemini, local)
Privacy (local option)	No	Yes (with Ollama)
Cost	$10/month	Free (pay for API tokens only)
Open source	No	Yes
IDE support	VS Code, JetBrains (paid)	VS Code, JetBrains (free)
Context window	Limited	Up to 200K (with Claude)

Installation

VS Code

Open VS Code Extensions (Ctrl+Shift+X)
Search “Continue”
Install the official extension by Continue
The Continue sidebar appears — click to configure

JetBrains

Settings → Plugins → Marketplace
Search “Continue”
Install and restart

Configuring Your LLM

Continue stores configuration in ~/.continue/config.json. Open it from the Continue sidebar → gear icon.

Using Claude (Anthropic)

{
  "models": [
    {
      "title": "Claude Sonnet 4.5",
      "provider": "anthropic",
      "model": "claude-sonnet-4-5",
      "apiKey": "sk-ant-YOUR_API_KEY"
    }
  ]
}

Claude’s 200K context window makes it particularly useful for large codebase analysis.

Using GPT-4o (OpenAI)

{
  "models": [
    {
      "title": "GPT-4o",
      "provider": "openai",
      "model": "gpt-4o",
      "apiKey": "sk-YOUR_OPENAI_KEY"
    }
  ]
}

Using Local Ollama (Free, Private)

First ensure Ollama is running with a model:

ollama pull codellama:13b
ollama serve

{
  "models": [
    {
      "title": "CodeLlama 13B (Local)",
      "provider": "ollama",
      "model": "codellama:13b"
    }
  ]
}

Multi-Model Configuration

Use different models for different tasks:

{
  "models": [
    {
      "title": "Claude - Chat & Complex Tasks",
      "provider": "anthropic",
      "model": "claude-sonnet-4-5",
      "apiKey": "sk-ant-..."
    },
    {
      "title": "CodeLlama - Quick Autocomplete",
      "provider": "ollama",
      "model": "codellama:7b"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Autocomplete",
    "provider": "ollama",
    "model": "starcoder2:3b"
  }
}

Use the free local model for autocomplete (many requests) and Claude for complex tasks.

Core Features

Open the chat panel and ask questions about your code:

Select code → Ctrl+Shift+L → “Explain this function”
“What does this regex pattern match?”
“Find potential bugs in this class”
“Suggest a more efficient approach”

Inline Edit (Ctrl+I)

Trigger inline AI editing at the cursor:

Place cursor in function you want to modify
Press Ctrl+I
Describe the change: “Add input validation and error handling”
Review diff → Accept or Reject

Context References with @

Use @ to add context to your questions:

@file — Reference specific files
@folder — Include entire directory
@codebase — Semantic search of your whole project
@terminal — Include recent terminal output
@problems — Include current VS Code errors/warnings
@docs — Search documentation (configure doc sources)

Example:

@file:src/auth.ts How can I make this authentication module support OAuth2 in addition to the current username/password system?

Autocomplete

Continue provides GitHub Copilot-style inline suggestions:

Enabled automatically for configured autocomplete model
Press Tab to accept, Esc to dismiss
Uses a fast local model (starcoder2, deepseek-coder) for low latency

Practical Workflows

Codebase Understanding

For a new codebase:

@codebase Explain the overall architecture and how data flows from the API endpoints to the database

Bug Investigation

With an error in the terminal:

@terminal @problems The tests are failing with this error. What's the cause and how do I fix it?

Refactoring with Full Context

When a refactor spans multiple files:

@file:services/user.ts @file:models/user.ts @file:routes/users.ts 
Refactor the user service to use dependency injection instead of directly importing the model

Documentation Generation

@file:utils/crypto.ts Generate JSDoc comments for all exported functions

Custom Slash Commands

Add custom commands to config.json:

{
  "slashCommands": [
    {
      "name": "test",
      "description": "Write unit tests for highlighted code",
      "prompt": "Write comprehensive unit tests for the following code using the project's testing framework. Include edge cases and error conditions:

{{{ input }}}"
    },
    {
      "name": "review",
      "description": "Code review for security and best practices",
      "prompt": "Review this code for security vulnerabilities, performance issues, and adherence to best practices. Be specific and provide code examples for suggested improvements:

{{{ input }}}"
    }
  ]
}

Invoke with /test or /review in the chat.

Cost Optimization

Continue usage costs depend on how much code context you send to API models:

Small questions with code selection: ~$0.001-0.005 per query (Claude Haiku)
Large codebase analysis: $0.01-0.10 per query (Claude Sonnet)

Optimization: Use local Ollama models for autocomplete and frequent simple questions; reserve API models for complex architecture questions and multi-file refactors.

Continue.dev is the right choice for developers who value model flexibility, privacy, and open-source tooling. The ability to switch between Claude’s reasoning, GPT-4o’s capabilities, and completely free local models makes it uniquely adaptable to different projects and budgets.

Continue.dev: Open-Source AI Coding Assistant for VS Code