The host handles models and coding
Claude Code, Cursor, Codex, and Trae still handle browsing, tool calls, file edits, and runtime execution.
AI can write code. Super Dev helps teams ship it properly.
The terminal only handles onboarding and upgrade. The real research, the three core docs, approval gates, spec, implementation, and delivery stay inside the host. Super Dev is not another code generator. It pushes the host back onto a commercial path that can be researched, reviewed, and shipped.
- Only remember super-dev, super-dev update, and super-dev uninstall
- After install, copy the host first prompt and go back into the host
- The first pass is research first, then the three core docs, then approval
The homepage now teaches uv install and the super-dev onboarding path. The installer prints the recommended host, the standard-flow first prompt, the competition-flow first prompt, and the post-onboard self-check. After that, the terminal should get out of the way and daily work moves back into /super-dev, $super-dev, or super-dev: inside the host.
v2.4.0 brings the unified host matrix, project-first onboarding, host-specific first prompts and resume playbooks, dual-mode readiness, and a stricter screenshot-grade UI gate forward into onboarding and delivery.
Installing a Python package is not the product. The installer pulls the current host matrix onto one path: onboard, copy the host-specific first prompt, then go back into the host and start research.
- Host matrix
- Unified
- Install to first prompt
- 5 min
- Governed stages
- 9
- Required approval gates
- 2
Current host coverage
CodeBuddy CLI
Copilot CLI
Codex CLIDDroid CLI
Gemini CLIKKimi Code
Qoder CLIQQwen Code
AntigravityCodeBuddyCCodeBuddyCN
Qoder
TraeTTraeCN
CodexTTrae SOLO
TTrae SOLOCN
WWorkBuddy
CodeBuddy CLICopilot CLICodex CLIDDroid CLI
Gemini CLIKKimi Code
Qoder CLIQQwen Code
AntigravityCodeBuddyCCodeBuddyCN
QoderTraeTTraeCN
CodexTTrae SOLO
TTrae SOLOCN
WWorkBuddy
Why Super Dev
Keep the host coding. Let Super Dev straighten the workflow.
Once installed, the host follows one governed path from research to delivery.
Super Dev handles process governance
The host moves through research, the three core docs, approval, spec, implementation, quality, and delivery.
The output is auditable delivery assets
PRD, architecture, UI/UX, spec, runtime validation, quality reports, and delivery archives are written as artifacts.
Who It Is For
Built for people already using AI coding and tired of unreliable delivery.
It is for teams and individuals already shipping with AI and now trying to make the process predictable, reviewable, and repeatable.
Heavy AI coding individuals
You already use Claude Code, Cursor, Codex, or Trae a lot, and you want the output to be more stable, more complete, and closer to actual delivery.
Small teams and startup teams
You need to ship MVPs and test business viability fast, but you do not want the AI development process to drift or disappear into chat history.
Project teams standardizing host workflows
You want different members using different hosts to follow the same process discipline, quality standards, and delivery evidence model.
Use Cases
Four common ways teams actually use it.
Super Dev fits real project execution more than one-off demos. These scenarios show where it adds the most value.
Build an MVP from zero to one
Start with similar-product research, then generate the three core docs, the spec, and frontend validation to ship a demoable MVP fast.
Take over an existing project
Generate a repo map, inspect the dependency graph and impact radius, then rebuild constraints, regression guard, and docs before the host starts changing code.
Standardize AI development across a team
Different members can keep their preferred hosts while Super Dev standardizes triggers, protocol surfaces, quality gates, delivery outputs, and workflow control commands.
Pull the host back from direct coding into a delivery path
When the host wants to jump into implementation, Super Dev forces research, the three core docs, approval, and only then implementation, while bugfix work follows a dedicated lightweight patch path.
Pipeline
From requirement to delivery, the path stays fixed.
The trigger is short and the workflow is fixed. The host moves through research, documentation, approval, implementation, verification, and delivery.
Research and requirement hardening
Read local knowledge first, research similar products online, and close gaps in scope, edge cases, and acceptance criteria.
Three core docs and spec
Generate PRD, Architecture, and UI/UX first. After approval, create the spec and task breakdown.
Frontend first, then backend integration
Build the frontend and verify it runs before moving into backend implementation and end-to-end wiring.
Quality and delivery
UI review, red-team checks, delivery packaging, release rehearsal, and release readiness determine whether the work can actually ship.
Critical gates
Three-doc approval gate
Without explicit approval, the host cannot create the spec or start implementation.
Frontend runtime gate
Without a passing frontend runtime report, the host cannot continue into backend and later stages.
Delivery ready / rehearsal passed gate
If the delivery package is not ready or the rehearsal did not pass, the project is not done.
Host Support
The unified host matrix is not just support coverage. It tells you how to start immediately after onboarding.
The homepage keeps only the signals that matter first: which class a host belongs to, the first prompt to use, what Super Dev loads first, and what that host is best suited for. Detailed self-checks, official workflow checks, and dual-mode readiness stay in the installer and docs center.
CLI
CLI 12CLI hosts are best when you want the shortest install-to-first-prompt path and a clear terminal or app companion workflow.
Claude CodeCodeBuddy CLICopilot CLICodex CLICursor CLIDroid CLIGemini CLIKimi CodeKiro CLIOpenCodeQoder CLIQwen Code
IDE
IDE 9IDE hosts keep workspace context, rules, commands, and longer-running implementation loops together.
AntigravityCodeBuddyCodeBuddyCNCursorKiroQoderTraeTraeCNWindsurf
Desktop assistants
Desktop assistants 5Desktop assistants are better for project instructions, knowledge-heavy work, and continuity across broader conversations.
ClaudeCodexTrae SOLOTrae SOLOCNWorkBuddy
Host matrix
RecommendedCompatible
| Host | Group | First prompt | Loads first | Best for | Status |
|---|---|---|---|---|---|
| Claude Code | CLI | /super-dev 你的需求 | CLAUDE.md + settings + project/user skills + subagents | 最快进入 research 和三文档的旗舰 CLI 宿主 | Recommended |
| CodeBuddy CLI | CLI | /super-dev 你的需求 | CODEBUDDY.md + rules + commands + skills + agents | 已经在 CodeBuddy CLI 里做长流程任务协作 | Compatible |
| Copilot CLI | CLI | super-dev: 你的需求 | copilot-instructions + AGENTS.md + skills + agents | 规则和 instructions 驱动的文本入口 | Compatible |
| Codex CLI | CLI | CLI: $super-dev | AGENTS.md + official skills + CLI $skill entry | Codex CLI 主战场，显式 skill 入口最稳 | Recommended |
| Cursor CLI | CLI | super-dev: 你的需求 | AGENTS.md + .cursor/rules + native resume | Cursor CLI 当前目录直接续流程，不靠项目 slash | Compatible |
| Droid CLI | CLI | /super-dev 你的需求 | AGENTS.md + .factory/rules + .factory/skills (+ commands compatibility) | Factory 会话里最快进入标准流或 SEEAI | Recommended |
| Gemini CLI | CLI | /super-dev 你的需求 | GEMINI.md + settings + TOML commands | GEMINI.md 和 custom commands 驱动的 slash 宿主 | Compatible |
| Kimi Code | CLI | super-dev: · /skill:super-dev · /flow:super-dev | AGENTS.md + explicit /skill:/flow entries + native resume (+ skills enhancement) | 中文长流程项目，显式 /skill:/flow 和原生继续链 | Compatible |
| Kiro CLI | CLI | /super-dev · super-dev: | AGENTS.md + steering + skills + native resume | steering + skills 一起约束的 CLI 宿主 | Compatible |
| OpenCode | CLI | /super-dev 你的需求 | AGENTS.md + commands + skills (+ agents enhancement) | commands/skills 驱动的原生 slash 宿主 | Compatible |
| Qoder CLI | CLI | /super-dev 你的需求 | AGENTS.md + rules + commands + skills (+ agents enhancement) | rules + commands + skills 同时工作的 Qoder CLI | Compatible |
| Qwen Code | CLI | /super-dev 你的需求 | QWEN.md + commands + skills + checkpoint/restore | QWEN.md + commands 模型，适合先研究再进实现 | Compatible |
| Antigravity | IDE | /super-dev 你的需求 | recommended GEMINI.md + custom commands (+ workflows enhancement) | 偏实验型 IDE，适合验证 Gemini-style project context | Compatible |
| CodeBuddy | IDE | /super-dev 你的需求 | CODEBUDDY.md + rules + skills + workspace continuity | 工作区连续性强，适合团队项目迭代 | Compatible |
| CodeBuddyCN | IDE | /super-dev 你的需求 | CODEBUDDY.md + rules + skills + workspace continuity | 中文工作区协作与连续返工 | Compatible |
| Cursor | IDE | /super-dev 你的需求 | Agent Chat + AGENTS.md + rules (+ beta commands) | Agent Chat 里直接做 UI 和实现返工 | Compatible |
| Kiro | IDE | /super-dev 你的需求 | AGENTS.md + steering + skills + agent continuity | Kiro IDE 里保持 steering 和 agent continuity | Compatible |
| Qoder | IDE | /super-dev 你的需求 | AGENTS.md + rules + commands + skills (+ agents enhancement) | IDE 里 rules/commands/skills 一起拉齐 Qoder 工作区 | Compatible |
| Trae | IDE | super-dev: 你的需求 | recommended project context + compatibility rules + optional skills | 先用项目上下文稳定接管旧项目工作区 | Compatible |
| TraeCN | IDE | super-dev: 你的需求 | recommended CN workspace skills + /plan /spec model | 中文工作区 + /plan /spec 组合的文本入口 | Compatible |
| Windsurf | IDE | /super-dev 你的需求 | AGENTS.md + workflows + skills | workflow 型 IDE 宿主，适合长链任务编排 | Compatible |
| Claude | Desktop assistants | super-dev: | Projects + project instructions + project knowledge + desktop extensions/MCP | Project / Knowledge / MCP 驱动的桌面助手 | Compatible |
| Codex | Desktop assistants | App/Desktop: / → super-dev | AGENTS.md + skills + enabled App/Desktop skill entry | Codex App/Desktop 里最完整的旗舰宿主体验 | Recommended |
| Trae SOLO | Desktop assistants | /super-dev 你的需求 | recommended workspace rules + optional skills model | SOLO 工作区里快速接入并继续当前流程 | Compatible |
| Trae SOLOCN | Desktop assistants | super-dev: 你的需求 | recommended CN workspace MTC / Code + skills model | 中文 SOLO 工作区，适合比赛和短平快交付 | Compatible |
| WorkBuddy | Desktop assistants | super-dev: 你的需求 | recommended task workbench + skills + MCP model | 任务工作台驱动的桌面助手连续工作流 | Compatible |
Proof / Trust
Visible trust signals.
Open source, published packages, host coverage, local knowledge, host validation, spec scoring, quality gates, and delivery artifacts all appear directly on the page.
MIT open source and PyPI release
The code is visible, the install path is clear, and versions are traceable. This is not a black-box hosted platform.
Multi-host integration and validation
The same governance model installs into CLI and IDE hosts, while the Host Validation Center tracks prerequisites, runtime acceptance, and delivery readiness.
Local knowledge first
knowledge/ and knowledge bundles are reused in research, the three core docs, spec generation, quality, and delivery.
UI Review, Spec Quality, and Release Readiness
The work is not done when code is generated. It must pass runtime validation, quality gates, spec-quality scoring, and release checks.
Auditable delivery artifacts
Repo Map, Dependency Graph, Impact Analysis, Regression Guard, Proof Pack, and release artifacts are written to disk.
10-Expert Agent Collaboration
PM, Architect, UI/UX, Security, Code, DBA, QA, DevOps, and RCA experts each govern their stage, constraining host output to professional standards.
How it works
The homepage now teaches the shortest path instead of a pile of terminal commands.
The installer tells you which host to use, what the first prompt is, and whether that host is ready to start. After that, research, the three core docs, approval gates, preview validation, quality, and delivery evidence continue inside the host rather than through a pile of low-level commands.
Terminal
uv tool install super-dev
# open the host installer
super-dev
# the installer prints:
# - the recommended host
# - the standard-flow first prompt
# - the competition-flow first prompt
# - the post-onboard self-check
# - the official workflow checks
# the terminal stops here
# real work moves back into the hostFAQ
These answers cover positioning, integration, and workflow.
The host handles the model, browsing, tool calls, file edits, and runtime execution. Super Dev handles research order, the three core docs, approval gates, frontend runtime validation, quality gates, and delivery standards.
A prompt template only affects one conversation. Super Dev installs into host protocol surfaces and persists approval gates, runtime validation, quality gates, delivery state, and release readiness.
Spec tools mainly govern project specs. Super Dev governs the AI development process inside the host, covering research, the three core docs, spec, frontend, backend, testing, and delivery in one governed path.
Those hosts use super-dev: as the text trigger, then rely on AGENTS, rules, steering, or skills to interpret and execute the Super Dev flow. The user only needs to remember the trigger, not the underlying protocol details.
The terminal only onboards. Real development moves back into the host within five minutes.
The homepage no longer teaches another layer of low-level commands. Install Super Dev, run super-dev so the installer writes the project-level surfaces, then copy the host-specific first prompt and return to the session you already work in. Research, the three core docs, approval gates, preview validation, and delivery all continue from there.
1. uv tool install super-dev
2. Run super-dev and let the installer print the recommended host and first prompt
3. Go back into the host and start with /super-dev, $super-dev, or super-dev:
The homepage now teaches uv install and the super-dev onboarding path. Source installs, pinned versions, and uninstall details stay in the docs center.
View source on GitHub