Cline

AI coding tools are now too numerous to count, but most are closed-source subscriptions where the vendor picks the model and your code passes through their servers. Cline is a different shape: the extension is open source and free (Apache-2.0, 66,168 stars), and you bring the model.

It works agentically — you describe the goal and it reads the relevant files, writes code, runs commands, reads the errors and iterates. The critical design decision is that every action is shown to you and executes only after you approve it: which file it wants to change, what the change looks like as a diff, what command it wants to run in the terminal — all visible and all refusable. For letting an AI touch your codebase, that control matters.

Model connectivity is its greatest flexibility and the part that must be stated clearly:

• Cloud APIs give the best results, but that part costs money — billed per token, paid directly to the model provider. Cline itself charges nothing, but do not assume the whole arrangement is free. A complex task can consume a substantial number of tokens, so set a spending cap first.
• Local models via Ollama or LM Studio are free and offline, with no code leaving your machine. The trade-off is that local models are usually markedly weaker than frontier cloud models on complex programming work, making them better suited to simpler tasks and privacy-sensitive situations.

That is why we did not award the offline badge — the extension can run offline, but whether the whole system does depends on the model you attach.

Also worth mentioning: MCP support (letting it call external tools and data sources), the ability to read browser screenshots for front-end debugging, and direct file and folder references in context.

Who it suits: developers who want AI assistance without being locked into one product, and who want to control both the model choice and where their data goes.