Continue is an open-source VS Code extension that connects your editor to any LLM — including local models running via Ollama — to give you AI-assisted coding that never sends your code to an external server. Where Cursor and GitHub Copilot route everything through cloud APIs, Continue lets you use Qwen2.5-Coder, Phi-4-mini, or any Ollama model for inline autocomplete and chat without leaving VS Code. This guide covers the complete setup, configuration for best results, and the daily workflows that make Continue genuinely useful.
Why Continue with Ollama
The combination of Continue and Ollama solves a specific problem: you want AI coding assistance in VS Code but either cannot or do not want to send your code to cloud APIs. This matters in several situations — working on proprietary codebases where sending code to external servers is not permitted, working offline or on a slow connection, wanting to avoid the ongoing subscription cost of cloud coding assistants, or simply preferring that your code stays on your machine. Continue is the best-maintained open-source VS Code coding assistant, and Ollama is the easiest way to run local models. Together they give you a fully functional, fully private AI coding workflow.
Installing Continue
Install Continue from the VS Code Extensions marketplace — search for “Continue” and install the extension from Continue Dev. It is free and open-source. After installation, the Continue icon appears in the left sidebar and a new chat panel opens on the right side of the editor. On first launch, Continue prompts you to choose a provider — select Ollama from the list of options. If Ollama is already running on your machine, Continue detects it automatically and lists your downloaded models in the model picker at the bottom of the chat panel.
Configuring Continue for Ollama
Continue’s configuration lives in a JSON file at ~/.continue/config.json on Mac and Linux, or %USERPROFILE%.continueconfig.json on Windows. Open it by clicking the gear icon in the Continue panel, or via Command Palette → “Continue: Open config.json”. The file defines your models, their roles (chat vs autocomplete), and context providers.
A practical starting configuration uses two Ollama models: a larger one for the chat panel where you ask questions and get explanations, and a smaller one for inline autocomplete where speed matters more than depth. Qwen2.5-Coder-7B works well for chat and Qwen2.5-Coder-1.5B for autocomplete. Set the chat model in the models array with the ollama provider and your model name, then set the autocomplete model in the tabAutocompleteModel field. Both point to Ollama’s default local URL. If Ollama is running on a different machine on your network, replace localhost with that machine’s IP address.
Figure 1 — Continue + Ollama: recommended model setup
Inline Autocomplete
Once the autocomplete model is configured, Continue shows grey ghost text suggestions as you type — similar to GitHub Copilot. Press Tab to accept a suggestion, Escape to dismiss it, or keep typing to ignore it. The suggestions trigger after a short pause in your typing, sending the surrounding context (the lines before and after your cursor) to the local Ollama model and streaming back the completion.
Autocomplete quality depends heavily on context window and model size. The 1.5B model handles single-function completions well but may struggle with completions that require understanding the broader codebase structure. For better context, open related files in VS Code before working — Continue includes open editor tabs as context when generating completions. You can also tell Continue which files to index for context by running “Continue: Index codebase” from the Command Palette, which creates a local embedding index of your project that autocomplete and chat can query.
The Chat Panel Workflow
The Continue chat panel is where most of the productivity value comes from. Select any code in the editor and press Cmd+L (Mac) or Ctrl+L (Windows/Linux) to add it to the chat context, then ask a question about it. Common workflows: select a function and ask “what does this do”, select a failing test and ask “why is this failing”, select a code block and ask “refactor this to use async/await”, or type a comment describing what you want and ask Continue to implement it.
The @-mention system lets you add specific context: type @file to reference a specific file, @codebase to search the indexed codebase for relevant context, @docs to reference documentation you’ve added, or @terminal to include recent terminal output. For debugging, @terminal is particularly useful — copy an error message into the terminal, then use @terminal in the chat to ask Continue to diagnose it with the actual error text as context.
Context Providers Worth Enabling
Continue’s context providers extend what it can access. The most useful for local development: the codebase provider (indexes your project with local embeddings for semantic search), the docs provider (lets you add documentation URLs that Continue scrapes and indexes locally), the diff provider (includes your current git diff as context — useful for asking “review my changes”), and the open tabs provider (includes all currently open editor tabs). All of these run locally and never send data externally when using Ollama as the backend.
Continue vs Cursor vs GitHub Copilot
The practical comparison: Continue with Ollama is the right choice when privacy is non-negotiable or you want zero ongoing cost. Cursor is the right choice when you want the best possible code quality from frontier models and are comfortable with cloud API costs. GitHub Copilot is the right choice when you need the most seamless editor integration in VS Code or JetBrains and want an enterprise-supported solution. Continue occupies a specific niche that neither Cursor nor Copilot fills: a fully private, fully local, open-source coding assistant that integrates directly into VS Code without requiring a subscription or sending any data externally.
Troubleshooting Common Issues
If autocomplete is not triggering, check that tabAutocompleteModel is set correctly in config.json and that the model name matches exactly what ollama list shows. If the chat panel shows “no models available”, Ollama is not reachable — verify it is running with ollama ps and check the apiBase URL in your config. If completions are very slow, switch to a smaller autocomplete model or reduce the context window size in the model config. If Continue cannot index your codebase, check that the project folder is open in VS Code (not just individual files) and that you have enough disk space for the embedding index.
Frequently Asked Questions
- Does Continue work with models other than Ollama?
- Yes. Continue supports Anthropic, OpenAI, Gemini, Azure OpenAI, Mistral, and any OpenAI-compatible API endpoint alongside Ollama. You can configure multiple providers and switch between them in the model picker at the bottom of the chat panel.
- Can Continue index large codebases?
- Yes, but indexing takes time proportional to codebase size. For very large repositories, limit the index to the most relevant directories using the .continueignore file (same syntax as .gitignore). The index is stored locally and only needs to be rebuilt when files change.
- Is Continue free?
- Continue itself is free and open-source (Apache 2.0 licence). Running it with Ollama local models has no ongoing cost. If you connect it to cloud providers like Anthropic or OpenAI, you pay those providers’ standard API rates.
- How does Continue compare to the Cursor editor?
- Cursor is a full VS Code fork with deeper AI integration and access to frontier models via its subscription. Continue is an extension for standard VS Code that works best with local models for privacy. If you cannot use Cursor due to company policy on cloud AI tools, Continue with Ollama is the practical alternative.
Last updated: October 2026
Keeping Continue Updated
Continue releases updates frequently — new model support, improved context providers, and bug fixes appear regularly. VS Code notifies you of extension updates automatically; apply them as they arrive since the extension is under active development and recent versions have meaningfully improved autocomplete quality and codebase indexing speed. The config.json format is stable across minor versions, so updates rarely break existing configurations. Check the Continue changelog on GitHub or the VS Code extension page when a major version is released to see if any configuration options have changed or new context providers are available. The community Discord (linked from the Continue docs site) is the best place to find working configurations for specific Ollama models and to troubleshoot setup issues that the documentation does not cover.
Continue with Ollama is the most practical path to a fully private AI coding assistant in VS Code today. The combination of a well-maintained extension, the flexibility to use any Ollama model, and zero data leaving your machine makes it a strong choice for anyone who needs AI coding help but cannot or does not want to use cloud-connected alternatives. The setup takes under ten minutes, and the config.json approach means you can tune model selection and context precisely for your workflow without waiting for a product team to add the feature you need.