Claude Code

·

Claude Code is Anthropic's coding agent for repository-scale engineering work. It reads files, edits code, runs commands, and coordinates subagents from the terminal, supported IDEs, the Claude desktop app, or a browser. Its strongest fit is a Claude-centered workflow that needs configurable permissions, persistent project instructions, and integrations with development infrastructure.

Cline

·

Cline is an open-source coding agent that runs as an extension in VS Code and JetBrains. It works in a two-phase Plan/Act loop: it proposes a plan, waits for approval, then makes the change. It is model-agnostic and bring-your-own-key, so the tool itself is free and the cost is the model API you point at it. Every file write, terminal command, and browser action is shown before it runs, which suits teams that need an audit trail.

Codex

·

Codex is OpenAI's coding agent across the CLI, IDE extension, desktop app, web, SDK, and cloud. Local sessions work directly in your repository with configurable sandbox and approval policies. Cloud tasks clone the repository into isolated OpenAI-managed environments, making Codex useful for parallel jobs that can run while you work elsewhere.

Cursor

·

Cursor is a VS Code-based editor that combines tab completion, inline edits, repository-aware agents, and delegated cloud agents in one interface. It is the strongest fit here for developers who want AI embedded in a graphical editor rather than centered on a terminal. The tradeoff is a more complicated usage model: cost depends on the plan, selected model, included usage pools, and optional on-demand usage.

GitHub Copilot

·

GitHub Copilot is a coding assistant that spans the IDE, CLI, and GitHub.com. Alongside chat and agent mode, it can run a cloud agent that opens a pull request from an assigned task and provides automated code review. Third-party agents such as Claude Code and Codex can run inside it. Billing moved to a usage-based model in mid-2026: every plan carries a monthly allowance of GitHub AI Credits, and code completion remains unlimited on paid plans.

OpenCode

·

OpenCode is an open-source coding agent for developers who want control over models, providers, permissions, and workflow. Its terminal interface is the main draw, but desktop, web, and IDE clients are also available. It can inspect a repository, plan changes without editing, modify files, run commands, and connect to external tools through MCP.

Roo Code

·

Roo Code is an open-source VS Code agent that started as a fork of Cline and grew its own architecture. It uses role-scoped modes - Code, Architect, Ask, and Debug - each with its own system prompt, tool permissions, and optional model, so you can route cheap questions to a light model and hard implementation to a frontier one. Boomerang Tasks let an orchestrator spawn specialized subagents for parallel work.

Windsurf

·

Windsurf is a code editor built around Cascade, an agent that handles multi-file edits, plus tab completion and inline edits that stay free of the usage quota. It ships its own fast agent model, SWE-1.5, alongside access to frontier models from OpenAI, Anthropic, and Google. Usage is metered by daily and weekly quotas rather than a monthly credit pool, and paid plans can buy extra usage at API rates.

AI coding tools compared
Tool Type Runs on Pricing Best for
Claude Code
Cline
Codex
Cursor
GitHub Copilot
OpenCode
Roo Code
Windsurf

AI agents

Agents are different from tools: a tool waits for your input, an agent takes a goal and runs with it. Every writeup discloses the autonomy level and the safety surface up front.

AI agents compared
Agent Autonomy Runs on Pricing Best for
OpenClaw Human-in-the-loop Self-hosted, Linux, macOS Free and open source Web automation that needs human review
Hermes Configurable (human-in-the-loop to autonomous) Self-hosted, Linux, macOS, Windows Free and open-weight Self-hosted, open-weight deployments
Manus Autonomous (with optional human checkpoints) Web Free tier with usage limits Multi-step tasks that mix research and writing