Continue

Free

Continue is an open-source AI coding agent that plugs into VS Code and JetBrains. You bring your own language model—cloud or local—and it sits beside you while you work.

Continue sits in the coding assistants category. It is not a finished product you subscribe to. It is a framework you configure.

You pick your LLM provider. You route autocomplete to a fast cheap model and chat to a strong one. You set up custom agents for your team's code review rules.

Cursor acquired Continue in 2024. The open-source codebase is still free and still available. New development now happens under Cursor.

Open it if you want control over your AI setup. Skip it if you want something that just works out of the box.

It rewards specific users who need flexibility more than polish.

• Developers who prioritize open-source and flexibility: You can inspect, fork, and change how it works. No vendor holds your code or your data.
• Teams with privacy or compliance requirements: You run local models on your own hardware. Sensitive code never leaves your network.
• Users who want flexibility in choosing their LLM provider: Switch between OpenAI, Anthropic, Google, or self-hosted models without reconfiguring everything.
• Developers on a budget who can't justify $20/month: The core tool is free. You only pay for the tokens you use, and local models cost nothing.
• Teams willing to fork and maintain their own version: You have engineers who can patch bugs and add features when the upstream stops.
• Those comfortable managing their own model routing: You set which model handles autocomplete, chat, and embeddings. You own the tradeoffs between speed, cost, and quality.
• Engineers building structured AI workflows: The three-mode setup—Chat, Plan, Agent—lets you stage AI work before it touches your files.
• Developers wanting AI assistance with broader context than inline completion: The @codebase command lets you ask about architecture across your whole project.
• Teams needing model governance and standardized AI workflows: You enforce which models your whole team uses and how they use them.
• Privacy-conscious teams with air-gapped requirements: Run 100% offline with local models. No cloud dependency at all.

These are the parts worth the setup time.

• Model-agnostic routing: You are not locked into one provider. Swap models per task without reinstalling.
• Full-project context: Ask about your entire codebase. It reads across files, not just the line you are typing.
• Local-first option: Run with Ollama on your own machine. Work offline with no data leaving your laptop.
• Interactive diffs: See changes before you accept them. Pick which lines to keep and which to reject.
• Three-mode workflow: Chat for quick questions, Plan to explore before acting, Agent to execute. You choose how much control to give up.

Skip it when the burden outweighs the benefit.

• Not actively maintained: The original project is read-only. You rely on a fork or move to Cursor.
• Setup overhead: You configure models, API keys, and routing rules. It takes time before your first useful response.
• Rough edges in daily use: Autocomplete sometimes fails and needs a restart. Diffs occasionally apply to wrong lines.
• No autonomous task completion: It will not finish a ticket for you. It helps you work, not replace your work.
• Self-hosted quality gap: Local models on consumer hardware do not match cloud API results. You trade privacy for power.