What for: an autonomous AI engineer inside VS Code — it reads the repo, edits files and runs commands with any model you plug in.

What it is and who it's for

Cline is an open-source autonomous coding agent inside VS Code: it reads your repo, edits files, runs commands and supports MCP — while you plug in whatever model you like (Claude, GPT, or a local one via Ollama). Built for developers who want a powerful in-IDE agent without vendor lock-in and with transparent control over every action. Its standout is the human-in-the-loop mode: it shows you the diff and the plan before anything is applied.

Key features

Autonomous file edits and command execution, with each step confirmed by you.
Any model: API providers or local (Ollama / LM Studio).
MCP server support for extending the agent with tools.
Plan/Act modes: plan first, then execute.
Diff previews and context control (which files the agent can see).
Fully open-source and free (you only pay for model tokens).

Getting started in 5 minutes

In VS Code, open Extensions, search for Cline and install it.
In the extension settings, pick a model provider and paste your API key (or point it at a local Ollama).
Open the Cline panel, describe the task and approve the edits and commands it proposes.

When to use it and when not to

✅ Use it if you want a free open-source agent in VS Code with your choice of model.
✅ Use it if control matters — seeing the diff and the plan before anything is applied.
❌ Skip it → go with "Roo Code" if you need custom modes/roles and fine-grained control over agent behavior.
❌ Skip it → go with "Claude Code" if you prefer the terminal and a maximally autonomous pipeline with verification.

Honest pricing

The extension itself is free and open-source. You pay for tokens of whichever model you choose, billed by its provider; with a local model via Ollama it's free. Exact numbers are on the provider's site and do change.

Gotchas

Token costs on big tasks with a strong model can surprise you — keep an eye on spend.
Quality equals the quality of the model you pick; a weak model means a weak agent.
Large context needs rationing, otherwise it gets expensive and slow.
Approve commands deliberately — the agent can run something destructive.

🤖 Prompt booster

"I'm using Cline in VS Code on a project. Suggest which model fits my budget (cloud vs local via Ollama), how to set up Plan/Act mode for the task , which files to add to context and which MCP servers to connect. Give me the wording for the first task so the agent doesn't break neighboring code."