Compare frameworks & add-ons

Practical capabilities and requirements
What matters	Basic Memory ↗	Beads ↗	BMad Method ↗	cc-sdd ↗	Compound Engineering ↗	OpenSpec ↗	RTK ↗	Spec Kit ↗	Superpowers ↗	TDD Guard ↗
Purpose	Persistent Markdown notes and a searchable knowledge graph available through MCP.	Give coding agents a persistent dependency-aware issue tracker and ready-to-work queue.	Guide product, architecture, development, and testing decisions through reusable delivery workflows.	Carry approved specifications through task execution, test-first implementation, and reviewer checks.	Connect planning and review with written lessons that inform the next change.	Keep proposals, specifications, designs, and implementation tasks together in the repository.	Compress supported command output before it enters the coding agent’s context.	Turn requirements into a specification, implementation plan, and concrete task sequence.	A development workflow built from reusable planning, debugging, testing, and review skills.	A Claude Code hook that checks test-first development and blocks unsupported implementation steps.
Type	Add-on	Add-on	Framework	Framework	Framework	Framework	Add-on	Framework	Framework	Add-on
Helps with	memory, context	memory, planning	planning, testing, review	planning, testing, review, orchestration	planning, review, memory	planning, context	context	planning, review	planning, testing, review	testing, review
Works with	Claude Code, Codex, Cursor, GitHub Copilot	Claude Code, Codex, Factory Droid, Cursor, Mux	Claude Code	Claude Code, Codex, Cursor, GitHub Copilot, Windsurf, OpenCode, Gemini CLI, Antigravity, Qwen Code	Claude Code, Cursor, Codex, GitHub Copilot, Factory Droid, Qwen Code, OpenCode, Pi, Oh My Pi, Antigravity, Devin CLI, Grok Build CLI, Kimi Code, Cline	Amazon Q Developer, Antigravity, Auggie, IBM Bob Shell, Claude Code, Cline, Command Code, CodeArts, CodeBuddy, Codex, Devin Desktop, ForgeCode, Continue, CoStrict, Crush, Cursor, Factory Droid, Gemini CLI, GitHub Copilot, iFlow, Junie, Kilo Code, Kimi Code, Kiro, Lingma, MiniMax Code, Mistral Vibe, Oh My Pi, OpenCode, Pi, SourceCraft Code Assistant for VS Code, Qoder, Qwen Code, Rovo Dev CLI, Zoo Code, Trae, Zed Agent, ZCode	Claude Code, GitHub Copilot, Cursor, Gemini CLI, Codex, Windsurf, Cline, Zoo Code, OpenCode, Pi, Mistral Vibe, Kilo Code, Antigravity, Kimi Code, Factory Droid	Alquimia AI, Amp, Antigravity, Auggie, Claude Code, Cline, CodeBuddy, Codex, Command Code, Cursor, DeepSeek Harness, Devin CLI, Docker Agent, Factory Droid, Firebender, ForgeCode, Gemini CLI, GitHub Copilot, Goose, Grok Build CLI, IBM Bob Shell, Junie, Kilo Code, Kimi Code, Kiro CLI, Lingma, Mistral Vibe, Muse Code, Oh My Pi, OpenCode, Pi, Qoder, Qwen Code, Rovo Dev CLI, SHAI (OVHcloud), Tabnine CLI, Trae, ZCode, Zed Agent	Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Factory Droid, OpenCode, Pi, Antigravity, Devin CLI, Grok Build CLI, Kimi Code	Claude Code
Installation	MCP	CLI / hooks	CLI / skills	Skills	Plugin / skills	Skills / commands	Hooks, Rules, Plugin, Extension	CLI / skills	Plugin, Extension, Package	Plugin / hooks
Requirements	uv / uvx for local installation; local Markdown storage.	The bd CLI and its documented Dolt storage setup.	Node.js 20.12+, Python 3.10+, and uv for Python-backed skills.	Node.js / npx and the selected coding harness.	A supported plugin-capable harness; see host-specific requirements.	Node.js 20.19.0 or newer.	The RTK binary, available through Homebrew, Cargo, or upstream releases.	Python, uv, Git, and a supported coding agent.	An installed coding harness with its plugin or skills facility.	Node.js 22+ and a supported test runner.
When it fits	Keeping project decisions and research accessible across coding sessions.	Multi-session work where tasks, blockers, and ownership should survive context resets.	Work ranging from a clear change to a larger project needing explicit product and architecture context.	Long-running changes that need requirements, design, tasks, and implementation to stay connected.	Projects where repeated decisions and review findings should become reusable team knowledge.	Iterating on changes to an existing codebase while keeping requirements reviewable.	Workflows producing repetitive build, test, package-manager, or Git output.	Teams wanting explicit project principles and versioned planning artifacts before implementation.	Projects that benefit from explicit design approval and a repeatable implementation process.	Existing Claude Code projects that want test results to gate file changes.
Limitations	MCP access is required. Automatic session briefings require the additional Claude Code plugin; other clients have different integration surfaces.	Adds a local database and task-management workflow. Agent setup may write instructions and hooks into the project.	Python-backed skills need uv. The installer offers additional integrations; only the harness setup explicitly verified here is listed.	Claude Code and Codex integrations are stable; the other skills integrations are beta. Qwen uses the legacy command mode.	Some workflows use multiple agents and additional model calls. Pi requires its subagent companion package for those workflows; host command syntax differs.	Command names vary by harness. Codex uses skills only; cross-repository stores remain beta.	Compression can hide detail; inspect raw output when needed. Some harnesses use instructions rather than interception; Copilot CLI uses deny-and-suggest.	Multiple integrations have compatibility checks. Pi needs an additional MCP extension for task-to-issue functionality.	Workflow guidance still depends on the agent following it. Each harness needs its own installation; companion subagent tools are optional on Pi.	Requires a supported test framework and reporter setup. Upstream now recommends its successor, Probity, for new projects; TDD Guard remains maintained.
Verified	2026-09-04	2026-09-04	2026-09-04	2026-09-04	2026-09-04	2026-09-04	2026-09-04	2026-09-04	2026-09-04	2026-09-04

Find your fit.