SuperClaude Framework

SuperClaude Framework is an open-source meta-programming configuration framework built on top of Anthropic's Claude Code. It enhances Claude Code through behavioral instruction injection and component orchestration, turning it into a structured development platform with systematic workflow automation. The project is MIT-licensed, available on GitHub under the SuperClaude-Org organization, and installable via pip or pipx.

What It Is

SuperClaude Framework is a configuration layer — not a standalone AI model — that injects specialized commands, cognitive personas, and development methodologies into Claude Code sessions. It provides 30 slash commands covering the full development lifecycle, 20 specialized AI agents with domain expertise, 7 adaptive behavioral modes, and 8 optional MCP server integrations. The framework is explicitly not affiliated with or endorsed by Anthropic.

Command and Agent Architecture

The 30 /sc:* slash commands are organized into functional categories:

• Planning & Design: /brainstorm, /design, /estimate, /spec-panel
• Development: /implement, /build, /improve, /cleanup, /explain
• Testing & Quality: /test, /analyze, /troubleshoot, /reflect
• Documentation: /document, /help
• Version Control: /git
• Project Management: /pm, /task, /workflow
• Research & Analysis: /research, /business-panel
• Utilities: /agent, /spawn, /load, /save, /sc, and more

The 20 specialized agents include a PM Agent, Deep Research agent, Security engineer, and Frontend architect, each providing domain-specific expertise on demand with automatic coordination based on context.

Behavioral Modes and Deep Research

Seven adaptive behavioral modes shape how Claude Code responds in different contexts: Brainstorming, Business Panel, Deep Research, Orchestration, Token-Efficiency, Task Management, and Introspection. The Deep Research mode, introduced in v4.2, enables autonomous multi-hop web research with up to 5 iterative search hops, confidence-based quality scoring (targeting 0.8 on a 0.0–1.0 scale), and case-based cross-session learning. Research depth levels range from Quick (5–10 sources, ~2 min) to Exhaustive (40+ sources, ~10 min).

MCP Server Integration

SuperClaude optionally integrates 8 MCP servers for enhanced performance. The README states that with MCP servers enabled, execution can be 2–3x faster with 30–50% fewer tokens compared to running without them:

• Tavily — Primary web search for Deep Research
• Context7 — Official documentation lookup
• Sequential-Thinking — Multi-step reasoning
• Serena — Session persistence and memory
• Playwright — Cross-browser automation
• Magic — UI component generation
• Morphllm-Fast-Apply — Context-aware code modifications
• Chrome DevTools — Performance analysis

Setup Path

Installation is straightforward via pipx (recommended) or pip:

pipx install superclaude
superclaude install

A direct Git clone with ./install.sh is also supported. After installation, restarting Claude Code activates all 30 commands. MCP server installation is optional and handled through the superclaude mcp CLI subcommand. A TypeScript plugin system is planned for v5.0 but is not yet available.

Update: v4.3.0 — Implementation Fixes, Security Hardening & Claude Code Alignment

The current stable release is v4.3.0, published in March 2026. The v4.x series stabilized the slash command architecture, expanded agents from 14 to 20, expanded modes from 6 to 7, and grew the command set from 22 to 30. The repository shows active development with 23,000+ GitHub stars and nearly 2,000 forks as reported by GitHub metadata. A v5.0 TypeScript plugin system is in active development (tracked in issue #419) with no set ETA. The project also maintains sibling frameworks — SuperGemini and SuperQwen — for other AI models.