Continue: An Open-Source Coding Assistant You Control

The question I get from clients about AI coding assistants is never about quality. It is "where does our code go, and who else sees it?"

Continue is the answer that lets you choose. It is an open-source assistant for VS Code and JetBrains where you configure which model handles which task — including models that never leave your network.

What you configure

The whole thing is a config file, which is the point:

{
  "models": [
    { "title": "Local", "provider": "ollama", "model": "qwen2.5-coder:7b" },
    { "title": "Cloud", "provider": "anthropic", "model": "claude-sonnet-4-5" }
  ],
  "tabAutocompleteModel": {
    "provider": "ollama", "model": "qwen2.5-coder:1.5b"
  }
}

Note the split. Autocomplete fires constantly and needs to be fast and cheap, so it runs locally on a small model. Chat and refactoring go to a stronger model when the work justifies it.

Why this matters commercially

I work with clients in security and government-adjacent sectors. For some of them, sending source code to a third-party API is not a preference to discuss — it is a contractual no.

Before tools like this, that meant no assistant at all. Now it means autocomplete and chat run against a model on their own hardware, and the productivity gap between those clients and everyone else mostly disappears.

Being able to answer the data-residency question in one sentence has won me work.

Practical setup notes

Use a model trained for code, not a general chat model — the difference in completion quality is large and immediate.

Size the autocomplete model down aggressively. A 1.5B model that responds in 200ms is far more useful than a 7B model that responds in two seconds, because slow autocomplete is worse than none.

Index the repository so chat can answer questions with your actual code as context rather than generic knowledge.

And add a .continueignore for anything sensitive. Even with a local model, keeping .env files out of context is basic hygiene.

The catch

It is more configuration than a commercial product where everything is decided for you. That is the trade you are making: some setup, in exchange for control over where your code goes.

Resources

• Repo: continuedev/continue
• Docs: docs.continue.dev
• Video walkthroughs: YouTube: Continue.dev setup
• Local backend: Ollama