Over the past two years large language models have expanded context windows and gained tool‑use abilities, turning AI-assisted coding from simple autocomplete into a genuine collaborator. Developers now see assistants that can read repositories, run tests, and browse documentation, reshaping how software is built and prompting a flood of new products.

The space splits along three axes: how users interact, where the code runs, and the level of autonomy. Terminal agents like Claude Code or Gemini CLI operate directly in the shell, while AI-native IDEs such as Cursor, Windsurf, Antigravity, and Zed embed assistants in the editor. Traditional IDE plugins—GitHub Copilot, Amazon CodeWhisperer, JetBrains AI Assistant, Replit Ghostwriter—offer low‑friction inline suggestions.

Pricing ranges from free tiers to credit‑based enterprise plans, while security and data‑privacy concerns push some teams toward on‑prem models. Choosing a tool hinges on role, workflow tolerance, and governance needs; solo engineers may adopt AI-native IDEs, whereas large firms often start with vetted IDE plugins before experimenting with higher‑autonomy agents.