A Detailed Comparison of Claude Code, OpenCode, Cursor, and GitHub Copilot for AI-Powered Coding

Maximilian Schwarzmüller

Introduction to AI-Powered Coding Tools [0:00]

The video provides an overview and comparison of popular AI tools for agentic engineering, focusing on Claude Code, OpenCode, Cursor, and Visual Studio Code with GitHub Copilot. The speaker highlights his extensive experience with these tools, having used them for various projects and courses.

Tool Categories and Interfaces [0:58]

The tools can be broadly categorized based on their primary interface:

• CLIs/TUIs (Command Line Interface/Terminal User Interface) [1:06]
  • Claude Code and OpenCode: These are primarily designed for terminal interaction.
  • OpenCode's Web Mode: OpenCode also has a web-based user interface, but the default and commonly used mode is the TUI.
  • IDE Integrations: Both Claude Code and OpenCode offer IDE integrations to preview changes or interact within an IDE, but they are optimized for command-line use.
  • Higher Flexibility: This approach provides developers with more flexibility, allowing them to focus on building effective AI agents rather than managing IDE complexities.
  • No Auto-completion: A drawback of CLIs/TUIs is the lack of AI-powered auto-completion in the editor.
  • Agentic Tools: Interaction is primarily through a chat interface that triggers AI actions.
• IDEs (Integrated Development Environments) [1:10]
  • Cursor and VS Code with GitHub Copilot: These are full-fledged IDEs (VS Code forks in the case of Cursor) that offer more than just AI agents.
  • Auto-completion: - Cursor: Provides "amazing" auto-completion that allows for rapid code editing across files. - GitHub Copilot: Offers "decent" auto-completion, though the speaker considers it less effective than Cursor's.
  • Agentic Tools: Like their CLI counterparts, these IDEs also offer chat interfaces to trigger AI, in addition to auto-completion.

Supported AI Models [4:24]

The availability and integration of AI models are crucial for these tools:

• Claude Code:
  • Primarily supports Anthropic's models: Haiku, Sonnet, and Opus.
  • The integration is noted as excellent due to Claude Code being from the same company (Anthropic).
  • Recently added Olama support, allowing users to run open-source models locally (if system resources permit), making it "free" but potentially less capable than advanced models like Opus.
• OpenCode, Cursor, and VS Code with GitHub Copilot:
  • Support Anthropic models (Haiku, Sonnet, Opus) as well, but they run within the tools' respective "harnesses."
  • Offer Olama support.
  • Support "lots of other models" from various providers and subscriptions.
  • OpenCode allows selection from a long list of models.
  • The speaker specifically praises the GPT-5.2 Codex model for coding, which is available in OpenCode, Cursor, and GitHub Copilot but not Claude Code.

Subscription Models and Open Source [8:12]

• OpenCode:
  • Is open source, promoting transparent, community-driven development and rapid innovation.
  • Offers optional paid subscriptions ("OpenCode Black") with different tiers, providing included usage.
  • Needs to be profitable, unlike larger, subsidized competitors.
  • Previously allowed using Anthropic models via Claude Code Max subscriptions, but Anthropic has since restricted this.
  • Can integrate with a GitHub Copilot subscription to access its models.
• Claude Code, Cursor, and GitHub Copilot:
  • All require paid subscriptions.
  • These tools are "heavily subsidized" by their respective companies (Anthropic, VC-funded Cursor, Microsoft for GitHub Copilot), potentially offering more usage per plan.

Core Features and Standards [10:18]

Most core agentic engineering features are widely supported across all tools, with new features quickly adopted by competitors.

• Agents and Skills:
  • Supported by all tools.
  • Claude Code: An "agent" typically refers to a sub-agent invoked for specific tasks (e.g., "explore"). It also has different "modes."
  • OpenCode: "Agents" comprise different operational modes (e.g., "plan mode," "edit mode") that users can cycle through, and specific agents for outsourcing work.
  • Standardization Issue: While features exist, the way they are configured and stored differs.
  • Rules Files/Memory: All tools support memory and rules files for persistent instructions and learnings across chat sessions. - Claude Code: Uses claude.md and a rules directory. - OpenCode: Supports agents.md (an open standard) and claude.md. - Cursor: Supports agents.md and Cursor rules. - GitHub Copilot: Supports agents.md and GitHub Copilot rules. - Incompatibility: The differing formats (e.g., claude.md vs. agents.md) create compatibility issues between tools.
  • Agent Skills Storage: Skills must be stored in specific, tool-dependent folders (e.g., opencodes/skills, claude/skills, cursor/skills, .github/skills). Claude Skills is the only format that works across all tools.
  • MCP Servers and Web Fetch: All tools support MCP servers (e.g., Playwright for browser access) and web fetch capabilities for sending web requests and reading documentation.

Verdict and Recommendations [15:32]

• Overall Quality: All tools are "decent" and produce good results depending on usage. There is no single "worst" tool.
• Claude Code: Excels with Anthropic models, especially Opus, due to native integration.
• OpenCode, Cursor, GitHub Copilot: Valuable for access to a wider range of models, such as GPT-5.2 Codex.
• IDE Preference: The speaker sometimes prefers IDEs (Cursor, GitHub Copilot) for tasks requiring more manual edits and a clear "diff view" to accept or deny changes.
• Developer's Role: The quality of AI output heavily relies on the developer's ability to provide good prompts, context, and effectively use features like agents and skills. Reviewing AI-generated solutions remains a critical developer task.