AI Coding Tools
10 tools listed in this category, compiled by hand from each provider’s official site and documentation.
AI coding tools operate at three levels of ambition. Completion tools suggest the next line or block while you type inside your existing editor, and they change nothing about how you work. Agentic assistants take an instruction, read across multiple files, and propose or apply a set of edits — closer to delegating a task than to autocomplete. App builders go further still and generate a running application from a description, which suits prototypes and internal tools more than it suits an existing production codebase.
The deciding factor for most developers is fit with the code they already have. A completion tool drops into any repository. An agent is only as good as its ability to understand the surrounding project, which degrades on large, unusual or poorly documented codebases. A builder that scaffolds its own stack is excellent from a blank slate and awkward when it has to respect decisions someone else made three years ago.
Two practical considerations apply to all of them. The first is code review: generated code compiles and looks plausible far more often than it is correct, so it needs the same scrutiny as a contribution from a new colleague, particularly around authentication, permissions and input handling. The second is data policy. Check whether your source is sent to a provider, retained, or used for training, and whether a self-hosted or zero-retention option exists if your employer requires one.
Reasonable starting points from this list: Cursor is the most popular full editor with agentic features and has a free tier. GitHub Copilot, paid, remains the default completion tool and integrates with the widest range of editors. Aider is free and open source, runs in the terminal against your own repository, and is worth trying if you prefer keeping your existing editor.
