Continue is a free, open-source AI coding assistant extension for VS Code and JetBrains, developed by Continue.dev and released under the MIT license. Unlike Copilot or Cursor, Continue does not provide AI itself — instead, it acts as a flexible IDE layer that connects to any model you choose: Claude via API, GPT-4o, Gemini, Mistral, or local models running through Ollama or LM Studio. This makes it the most cost-controllable option available: the extension is free, and you pay only for the API calls you make (or nothing, if you run a local model). The key differentiator over GitHub Copilot is full model choice, full data control, and zero subscription lock-in.
What It Does
Continue adds a chat panel and inline editing commands to your existing VS Code or JetBrains IDE. You configure which model to use per task — for example, a fast local model for autocomplete and a larger cloud model for complex reasoning. The @codebase command indexes your project so the model can answer questions with full file context. Inline editing lets you highlight code and describe the change you want. Continue also supports custom slash commands (reusable prompts), context providers that pull in documentation, GitHub issues, or web search results, and multi-file diff review. All configuration lives in a local JSON file, and the extension never stores your code on Continue’s servers.
Who It’s For
- Developers who want BYOK (bring your own key) to avoid per-seat subscription costs
- Privacy-conscious developers who need to ensure code never reaches a third-party server, achievable by using local models via Ollama
- Experimenters and power users who want to swap models freely to compare quality
- Teams on a budget who have API access through other agreements (e.g., AWS Bedrock, Azure OpenAI)
- Open-source contributors who want a fully auditable, hackable coding assistant
- Developers in JetBrains IDEs who lack the Cursor option and find Copilot’s JetBrains integration weaker
Pricing
| Plan | Price | Key Limits |
|---|---|---|
| Extension | Free | MIT open-source, no usage limits |
| Model costs | Pay-per-use | You pay your own API provider directly |
| Local models | $0 | Use Ollama or LM Studio — no API costs |
Verification note: Continue extension is free and open-source. Model costs depend on your chosen provider. Verified at continue.dev as of 2026-04-14.
Key Features
- Any model, any provider: connect to Anthropic, OpenAI, Google, Mistral, Ollama, LM Studio, AWS Bedrock, Azure OpenAI, or self-hosted models
- Local model support: run completions fully on-device via Ollama with no data leaving your machine
- @codebase context: indexes your project and retrieves relevant files to answer architecture and usage questions
- Custom slash commands: define reusable prompts (e.g.,
/test,/explain,/document) in your config - Context providers: pull in docs, GitHub issues, terminal output, or web pages alongside your code context
- Multi-file diffs: review proposed changes across files before accepting
- Fully open-source (MIT): inspect, fork, or contribute to the codebase at github.com/continuedev/continue
Limitations
- No AI included. You must configure and pay for your own model. This is a feature for power users but a friction point for developers who just want something that works out of the box.
- Autocomplete quality depends on your model choice. A misconfigured or slow local model produces poor completions. Matching model to task takes experimentation.
- No agent mode. Continue does not have a Composer/agent mode for autonomous multi-file implementation. It is a chat-and-inline-edit tool, not an agentic one.
- Setup overhead. First-time configuration (API keys, model selection, context providers) takes more effort than installing Copilot.
- Moat is low. Continue’s value is in the plugin layer, not proprietary AI. Competitors can (and do) add BYOK options, reducing Continue’s uniqueness over time.
Bottom Line
Continue earns a perfect 10/10 value score — it is completely free — and solid utility (7/10) when configured well. The moat is low (5/10) because it competes purely on openness and flexibility, not proprietary capability. It is the best option for developers who want zero subscription cost, full model flexibility, or full data control. For developers who want the best out-of-the-box experience without configuration work, GitHub Copilot at $10/mo is easier and more capable.
Best Alternatives
| Tool | Price | Key Difference |
|---|---|---|
| GitHub Copilot | $10/mo | Better out-of-box experience, chat + agent mode included |
| Cursor | $20/mo | Full AI IDE, agent mode, no setup required |
| Codeium | Free | Also free, but closed-source with included AI |
| Aider | Free | CLI-based BYOK tool with stronger agentic capability |
FAQ
Does Continue cost anything? The Continue extension is completely free and open-source under the MIT license. You pay only for model API usage — or nothing if you use local models via Ollama or LM Studio.
Can I use Continue without sending code to the cloud? Yes. Configure Continue to use a local model through Ollama or LM Studio and your code never leaves your machine. This is one of Continue’s strongest use cases.
How does Continue compare to GitHub Copilot? Copilot is easier to set up and more capable out of the box. Continue is free and gives you full control over which model you use. If you already pay for Claude or OpenAI API access, Continue can be the cheaper option. If you want zero configuration, Copilot wins.
Sources
- Official website — verified 2026-04-14
- GitHub repository — verified 2026-04-14
- Continue documentation — verified 2026-04-14