Authority AI Editorial
Cline: an autonomous coding agent inside VS Code
Cline is an open-source autonomous coding agent that runs primarily as a VS Code extension, though the project also ships a JetBrains plugin, a desktop app, and a command-line tool. It is for developers who want an agent that can read a codebase, propose an edit, run a shell command, and check its own output, while keeping a human in the approval loop by default rather than making changes silently.
What it does
Cline reads and edits files across a project, executes terminal commands, and watches the resulting output for errors so it can correct course. It works in two modes. In Plan mode, Cline's own documentation says it "can read your codebase, run searches, and discuss strategy, but cannot modify any files or execute commands," which keeps early exploration and architectural discussion separate from anything that touches disk. Once you are confident in the approach, you switch to Act mode, where Cline "retains the full context from your planning session and can now modify files, run commands, and execute your strategy," carrying the conversation forward rather than starting over. You can cycle back to Plan mode mid-task if a change turns out to be more involved than expected. Every file edit and command runs past you for approval unless you turn on auto-approve for a session, which is the mechanism the project uses to keep an agent that can run arbitrary shell commands from doing so unsupervised. Project-specific instructions live in a .clinerules file, and the agent can be extended with Model Context Protocol servers for the tools and data sources it needs.
How you use it
Most people install Cline from the VS Code Marketplace or the JetBrains plugin marketplace and work inside their existing editor. You give it a task in plain language, watch it propose a plan in Plan mode, then switch to Act mode to let it execute, approving edits and commands as they come up. A typical session starts by describing the goal, letting Cline explore the relevant files and raise questions or edge cases, discussing the approach until you are satisfied, and only then switching modes to let it write and run code. The desktop app and CLI extend the same agent outside an IDE, including a headless mode meant for CI pipelines. Cline's own documentation describes the desktop app as "open source for open weight models," meaning you connect your own model access rather than going through a bundled provider.
What to know before adopting it
Cline is licensed under Apache 2.0 and written primarily in TypeScript. As of September 2026 the project's GitHub page reports roughly 69,000 stars, which reflects adoption but is not a measure of code quality. The cost model is bring-your-own-key: you connect Anthropic, OpenAI, Google, a local model, or another supported provider and pay that provider directly, though the VS Code extension also offers a pay-as-you-go Cline provider and a subscription option called ClinePass if you would rather not manage separate API keys. The project is under active development, with the GitHub releases page showing separate, frequent version streams for the desktop app, the VS Code extension, and the CLI, each cut on its own schedule rather than a single unified version number. That split matters if you standardize on one entry point: a fix that lands in the VS Code extension does not automatically appear in the desktop app or CLI release the same week, so it is worth checking which surface a changelog entry actually applies to before assuming a bug is fixed everywhere.
Where it fits next to similar tools
Cline occupies the IDE-embedded end of the coding-agent spectrum, closer to a supervised pair programmer than a fire-and-forget script. That makes it a reasonable fit for teams that want agent-driven edits without giving up per-change review, in contrast to terminal-first tools that commit changes as they go or fully autonomous pipelines meant to run without a human watching each step. Anyone who wants the same written methods for coding agents across tools can also look at the AI Ultimate Skill Collection, 3,029 ready-made skills in 12 packs that your agent follows step by step.