AI Coding Agents

AI coding agents do more than suggest the next line. You give them a goal — “fix this failing test”, “migrate this module” — and they read the codebase, plan, edit files, run commands and tests, and often open a pull request. Some run in your terminal, others in cloud sandboxes that keep working after you close your laptop. This category also covers prompt-to-app builders, whose agents generate and host whole applications.

How to choose:

• Where it runs: a local terminal agent sees your environment; a cloud agent works in parallel but needs repository access and a sandbox you trust.
• Permissions and safety: look for approval modes, command allow-lists and clear audit trails before letting an agent run unattended.
• Model access and limits: check which models the agent can use and how usage limits or API costs behave in long sessions.
• Integration: IDE extensions, Git hosting and CI support decide how well an agent fits your existing workflow.
• Code ownership (app builders): check Git sync, export and how tied the app is to the builder’s hosting before an experiment becomes production.

Our reviews: Claude Code, OpenAI Codex, Replit, Lovable, Bolt.new, v0, Cline, Amp.