Continue

What is Continue?

Continue is a next-generation open-source AI software development copilot. The extension runs in your code editor (VS Code or JetBrains), providing advanced features similar to GitHub Copilot, but with total freedom for the developer to connect any language model (LLM) of their choice — whether through commercial APIs in the cloud (such as Claude 3.5 Sonnet, Gemini Pro, or GPT-4o) or by running local offline models at no cost at all.

How it works

After installing the Continue extension in your editor, you configure the JSON control file (config.json). In it, you define which language models will be used for two main tasks: Sidebar Chat (for answering questions and refactoring large code blocks) and Tab Autocomplete (which predicts the next lines you will type in real time). Continue reads the context of the open files in the project and responds to interactive commands initiated directly in the code editor.

Key features

• Contextual Sidebar Chat: An interactive tab to ask questions, explain code snippets, and plan architectures.
• Tab Autocomplete: Dynamic line-by-line code suggestions when pressing the Tab key.
• Project Codebase Reference: Use the /codebase command to ask the AI to scan your entire project folder, providing answers with a complete view of all classes and dependencies.
• Quick Documentation: Use the @docs command to integrate online library documentation directly into the AI's chat context.
• Local Model Support: Simple integration with Ollama and LM Studio to ensure absolute code privacy.

Available integrations

Continue operates in an integrated manner with all major APIs and local environments:

• Compatible IDEs: Microsoft VS Code and IDEs from the JetBrains family (such as IntelliJ, WebStorm, PyCharm).
• Model providers: Ollama, DeepSeek, Anthropic, Gemini, OpenAI, Together AI, Mistral AI, OpenRouter, and more.

Who it is for

• Individual Developers and Freelancers interested in a free, highly customizable, and fast copilot.
• Enterprise Software Engineers with strict code privacy restrictions who need to run AI models completely locally and offline.
• Engineering Teams who want to standardize the use of low-cost programming language models (such as DeepSeek Coder) to reduce operational costs.

Pricing

Pros and cons

Pros:

• Absolute freedom to choose the cheapest or smartest model on the market.
• Incomparable data privacy when running locally on your personal hardware.
• Easy to reference entire files using the @ symbol in the chat.

Cons:

• Requires basic knowledge of JSON configuration files to configure APIs and custom behaviors.
• The quality of real-time autocomplete depends directly on the computer's processing power if the model is run locally.

Alternatives to Continue

The main competing tools for copilots and code agents are: GitHub Copilot, Cursor Editor, Cline, and Tabnine.