Autonomous coding
Long-horizon, unattended coding sessions where an agent works for minutes or hours without human approval at each step. Contrasts with approval-gated modes where the user confirms each shell command or file edit.
Background
Autonomous coding pushes the agent loop further by allowing the model to make decisions and execute actions without per-step human confirmation. Used for refactors, test generation, dependency upgrades, and bug-bash runs. Requires strong safety primitives — sandboxed execution, file-system permissions, git rollbacks — because mistakes propagate quickly. Devin and OpenHands lead in long-horizon autonomous runs; in-editor agents like Cursor and Windsurf default to approval-gated modes.
Tools that use it
- 01→OpenHands
Open-source autonomous coding agent platform, formerly OpenDevin, by All Hands AI.
- 02→Devin
Cognition's autonomous software engineer — runs tasks in a managed cloud sandbox with browser, terminal, and editor.
- 03→Claude Code
Anthropic's official CLI agent for Claude — runs in the terminal, edits files, executes commands, and ships PRs.