Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
作者提供的说明(英文原文)
AI Constraint Engine — enforces CLAUDE.md and .cursorrules as laws. 51 MCP tools.
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"speclock": {
"command": "npx",
"args": [
"-y",
"speclock"
]
}
}
}在终端中运行这条命令
claude mcp add speclock -- npx -y speclock粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"speclock": {
"command": "npx",
"args": [
"-y",
"speclock"
]
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"speclock": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"speclock"
]
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
这个服务器在你的电脑上运行,而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。
你的选择通过 Cookie 保存在这台设备上,无需账户。
这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月6日
注册名称已通过 GitHub 账户 sgroy10 验证,该账户拥有此仓库
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 1 项适用的基础检查中通过了 1 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于–,最新版本(5.5.3)于7个月前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Node.js。
同一分类中有:Spec Driven Development, Remote Desktop Commander, dotnet-mcp
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
MCP server for .NET symbols, XAML, source generators, and restricted workspace edit
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
Rules files tell AI what not to change. SpecLock enforces them.
Stop Claude Code, Cursor, Codex, Windsurf, and other AI coding tools from crossing project constraints you already wrote in CLAUDE.md , AGENTS.md , and .cursorrules .
Install · See a block · Claude Code plugin · Website · Saves Wall
Why another rules tool? Rules files are context. They can be forgotten, diluted, or overridden during a long coding session. SpecLock turns those rules into checks that run before edits, shell commands, and commits.
CLAUDE.md: Never modify the authentication system.
You: Add social login to the login page.
Without SpecLock Claude: I'll update the auth flow and add an OAuth provider...
With SpecLock (strict mode) SpecLock: BLOCKED — conflicts with "Never modify the authentication system" Match: login → auth → authentication The action was denied before the files changed.
SpecLock uses semantic conflict detection rather than simple keyword matching. It catches indirect actions such as “clean up old patient data,” “streamline checkout,” or “temporarily disable MFA” when they violate an active constraint.
Run this from the project you want to protect:
npx speclock@latest protect # reads existing AI rule files; advisory by default
npx speclock@latest doctor # confirms rules, hooks, and integration
When the advisory output looks right, enable blocking:
npx speclock@latest protect --strict
No account is required. SpecLock runs locally by default, and advisory mode never blocks a change.
Inside Claude Code, run:
/plugin marketplace add sgroy10/speclock
/plugin install speclock@speclock-marketplace
Then protect the current project once:
npx speclock@latest protect # warnings only
npx speclock@latest protect --strict # block confirmed conflicts
The plugin automatically starts SpecLock's MCP server and checks Claude Code Write, Edit, and Bash actions before execution. It includes all 51 MCP tools and works alongside your existing CLAUDE.md.
SpecLock is packaged for multiple agent ecosystems, but the enforcement level depends on what each host exposes:
| Platform | Install/discovery path | Protection level |
|---|---|---|
| Claude Code | Native marketplace plugin above | Native pre-action checks for Write, Edit, and Bash |
| Gemini CLI | Install this repository as a Gemini extension | MCP-assisted checks plus project context |
| Cursor | Agent Plugin / Cursor marketplace package | MCP-assisted checks plus rules |
| Codex | Repository Codex plugin in plugins/speclock | MCP-assisted checks plus $speclock-guardrails skill |
| GitHub Copilot CLI | Add this repository as a plugin marketplace | MCP-assisted checks plus bundled plugin context |
| Cline | MCP server; curated marketplace submission in progress | MCP-assisted checks |
| Windsurf | speclock mcp install windsurf | MCP-assisted checks plus rules |
| Any Git client or CI | speclock protect | Commit/CI enforcement independent of the coding agent |
MCP-assisted means the agent can call SpecLock before acting; it does not guarantee interception. Use speclock protect --strict and CI when a constraint must be enforced regardless of the client.
SpecLock has a different job from memory and skills: memory recalls context, skills provide procedures, and SpecLock verifies planned actions against explicit constraints. It reduces constraint drift; it cannot guarantee factual correctness or make a model hallucination-free.
Repository installs supported by current clients:
gemini extensions install https://github.com/sgroy10/speclock
copilot plugin marketplace add sgroy10/speclock copilot plugin install speclock@speclock-marketplace
| Capability | What it does |
|---|---|
| Pre-action checks | Reviews Claude Code writes, edits, and shell commands before they run |
| Semantic constraints | Detects synonyms, euphemisms, compound requests, and indirect violations |
| Git enforcement | Adds a second guard at commit time |
| Advisory and strict modes | Start with warnings; opt in to hard blocking when ready |
| Audit trail | Records decisions in a tamper-evident HMAC chain |
| MCP integration | Exposes 51 tools to Claude Code, Cursor, Codex, Windsurf, and Cline |
| Shareable save receipts | Shows what SpecLock prevented with speclock wins |
speclock protect # extract constraints and install project protection
speclock protect --strict # enable hard enforcement speclock doctor # verify the complete setup speclock check "action description" # preview whether an action conflicts speclock add-lock "rule" # add a constraint explicitly speclock list-locks # inspect active constraints speclock enforce hard|advisory # switch enforcement mode speclock mcp install # Claude Code, Cursor, Windsurf, Cline, or Codex speclock wins # create a shareable save receipt
Full reference: npx speclock@latest help
$speclock-guardrails skill and pinned MCP server.More links and project badges
Smithery · Badge gallery · Sandeep Roy
AI coding tools have memory now. Claude Code has CLAUDE.md. Cursor has .cursorrules. Mem0 exists.
But memory without enforcement is useless.
Your AI *remembers* you use PostgreSQL — then switches to MongoDB because it "seemed better." Your AI *remembers* your auth setup — then rewrites it while "fixing" a bug. You said "never touch the payment logic" 3 sessions ago — the AI doesn't care.
Remembering is not respecting. No existing tool stops the AI from breaking what you locked.
You set constraints. SpecLock enforces them — across sessions, across tools, across teams.
speclock lock "Never modify auth files" → auto-guards src/auth/*.ts
speclock lock "Database must stay PostgreSQL" → catches "migrate to MongoDB" speclock lock "Never delete patient records" → catches "clean up old data" speclock lock "Don't touch the payment flow" → catches "streamline checkout"
The semantic engine doesn't do keyword matching. It understands:
And it knows what's safe:
npx speclock setup --goal "Build my app" --template nextjs
Creates SPECLOCK.md, injects rules into package.json, generates .speclock/context/latest.md. The AI reads these automatically.
Install the plugin using the commands above. For MCP-only setup without the plugin hooks, add this to .mcp.json:
{
"mcpServers": { "speclock": { "command": "npx", "args": ["--yes", "speclock@5.8.0", "serve", "--project", "."] } } }
Same config — add to .cursor/mcp.json or equivalent.
https://speclock-mcp-production.up.railway.app/mcp| Claude Memory | Mem0 | .cursorrules | SpecLock | |
|---|---|---|---|---|
| Remembers context | Yes | Yes | Manual | Yes |
| Blocks the AI from breaking things | No | No | No | Yes |
| Semantic conflict detection | No | No | No | Yes — covered by adversarial tests |
| Tamper-proof audit trail | No | No | No | HMAC-SHA256 chain |
| Hard enforcement (AI cannot proceed) | No | No | No | Yes |
| SOC 2 / HIPAA compliance exports | No | No | No | Yes |
| Encrypted storage (AES-256-GCM) | No | No | No | Yes |
| RBAC + API key auth | No | No | No | 4 roles |
| Policy-as-Code DSL | No | No | No | YAML rules |
| Works on Bolt.new, Lovable, etc. | No | No | No | Yes |
Other tools remember. SpecLock enforces.
Not keyword matching — semantic analysis with an optional Gemini Flash hybrid for grey-zone and cross-domain cases. The repository includes adversarial, false-positive, question-framing, patch-gateway, and diff-analysis test suites.
Category Detection Example Direct violations 100% "Delete the auth module" vs lock "Never modify auth" Euphemistic attacks 100% "Clean up old patient data" = deletion Temporal evasion 100% "Temporarily disable MFA" = disable MFA Dilution attacks 100% Violation buried in multi-part request Compound sentences 100% "
只显示了开头部分;其余内容请在仓库中阅读