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)
Draft, publish, and manage your VeloCMS blog from Claude, Cursor, or any MCP client.
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。
VELOCMS_SITE_URL, VELOCMS_API_KEY标为机密的值是你账户的钥匙。千万不要分享给别人,也不要粘贴到聊天里。
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": {
"velocms": {
"command": "npx",
"args": [
"-y",
"@velocms/mcp"
],
"env": {
"VELOCMS_SITE_URL": "your-value",
"VELOCMS_API_KEY": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add velocms --env VELOCMS_SITE_URL=your-value --env VELOCMS_API_KEY=your-value -- npx -y @velocms/mcp运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"velocms": {
"command": "npx",
"args": [
"-y",
"@velocms/mcp"
],
"env": {
"VELOCMS_SITE_URL": "your-value",
"VELOCMS_API_KEY": "your-value"
}
}
}
}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": {
"velocms": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@velocms/mcp"
],
"env": {
"VELOCMS_SITE_URL": "your-value",
"VELOCMS_API_KEY": "your-value"
}
}
}
}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7日
注册名称已通过域名 velocms.org 验证(注册中心会核实域名所有权)
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。
你需要填写:VELOCMS_SITE_URL, VELOCMS_API_KEY
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(0.1.3)于3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, AvaHR
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.
Read hiring data and create inactive assessment and questionnaire drafts for human review.
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。
An MCP server for VeloCMS — draft, edit, publish, and manage your blog (posts, media, comments, members, site settings) from Claude Code, Claude Desktop, Cursor, or any other MCP client.
It's a thin, typed wrapper around the VeloCMS Public API (https:// /api/v1, documented at /api/openapi on any VeloCMS site): every tool maps to one real API call, with Zod-validated inputs and readable errors — no scraping, no browser automation, just the platform's own supported REST API.
| Tool | What it does |
|---|---|
list_posts | List posts, paginated, optionally filtered by status (draft/published) |
get_post | Fetch a single post by ID, including its full content and SEO fields |
create_post | Create a new post (defaults to draft) |
update_post | Partially update an existing post — only the fields you pass change |
delete_post | Permanently delete a post |
publish_post | Set a post's status to published (stamps published_at) |
unpublish_post | Set a post's status back to draft |
list_media | List the media library, paginated, optionally filtered by MIME type |
list_comments | List comments, paginated, optionally filtered by post or moderation status |
moderate_comment | Set a comment's status to approved, pending, or spam |
list_members | List subscribers/readers (emails are masked, e.g. u***@example.com) |
get_site_settings | Read the blog's name, description, and feature flags |
Every tool ships a description an LLM can read to figure out when and how to use it — that's the point of MCP, not just a REST proxy with extra steps. See Tool reference below for full argument lists.
In your VeloCMS dashboard: Settings → API Keys → create a key with the scopes you need (posts:read, posts:write, media:read, comments:read, comments:moderate, members:read, site-settings:read, etc.). API access requires the Pro plan or higher.
You don't need to clone this repo — npx fetches and runs it on demand.
claude mcp add velocms \
--env VELOCMS_SITE_URL=https://myblog.velocms.org \ --env VELOCMS_API_KEY=velo_your_64_char_hex_key_here \ -- npx -y -p @velocms/mcp velocms-mcp
MCP servers register their tools at session start, so restart your Claude Code session after adding this.
Add to your claude_desktop_config.json (Settings → Developer → Edit Config):
{
"mcpServers": { "velocms": { "command": "npx", "args": ["-y", "-p", "@velocms/mcp", "velocms-mcp"], "env": { "VELOCMS_SITE_URL": "https://myblog.velocms.org", "VELOCMS_API_KEY": "velo_your_64_char_hex_key_here" } } } }
Restart Claude Desktop after editing.
Same shape as above — point the client's MCP config at npx -y -p @velocms/mcp velocms-mcp with VELOCMS_SITE_URL and VELOCMS_API_KEY set in its env. The server speaks standard MCP over stdio, so any client that supports stdio MCP servers works.
git clone https://github.com/VeloCMS/velocms-mcp.git
cd velocms-mcp npm install npm run build cp .env.example .env
VELOCMS_SITE_URL=... VELOCMS_API_KEY=... node dist/index.js
| Env var | Required | Description |
|---|---|---|
VELOCMS_SITE_URL | yes | Your blog's base URL — a *.velocms.org subdomain or a bound custom domain. No trailing slash needed. |
VELOCMS_API_KEY | yes | An API key from /admin/settings → API Keys. Requires Pro plan or higher. Never logged. |
If either is missing, the server prints a clear message to stderr and exits immediately (process.exit(1)) — it never starts half-configured.
Argument names are camelCase; the server maps them to the API's snake_case JSON fields for you.
`list_posts` — page?, perPage? (max 100), status? (draft | published).
`get_post` — id (required).
`create_post` — title (required, ≤255 chars), slug?, contentHtml?, contentJson? (TipTap ProseMirror document — prefer contentHtml unless you need this), excerpt? (≤500), status? (draft default | published), tags? (string array), seoTitle? (≤60), seoDescription? (≤160).
`update_post` — id (required) + any of the create_post fields (all optional here). At least one field besides id is required — the tool rejects a no-op call before making any network request.
`delete_post` — id (required). Permanent.
`publish_post` / `unpublish_post` — id (required). Shorthand for update_post({ status: "published" }) / update_post({ status: "draft" }).
`list_media` — page?, perPage?, type? (MIME type prefix, e.g. "image").
`list_comments` — page?, perPage?, postId?, status? (approved | pending | spam).
`moderate_comment` — id (required), status (required: approved | pending | spam).
`list_members` — page?, perPage?, tier? (free | paid).
`get_site_settings` — no arguments.
The VeloCMS API returns a consistent JSON error envelope:
{ "error": { "code": "RATE_LIMITED", "message": "...", "details": {} } }
This client surfaces that message directly (never a raw stack trace), enriched with actionable hints:
UNAUTHORIZED) — the message tells you to check VELOCMS_API_KEY.PLAN_UPGRADE_REQUIRED) — the message points at /admin/billing.INVALID_SCOPE/FORBIDDEN) — the message tells you to check the key's scopes.RATE_LIMITED) — the Retry-After response header (seconds) is parsed and included in the error message, and returned as retryAfterSeconds if you're calling the client library directly.code + message are surfaced as-is.Rate limits are plan-based (Pro: 30/min, 1,000/hr · Business: 120/min, 5,000/hr · Agency: 300/min, 20,000/hr) — this server does not retry automatically; it fails fast with the wait time so an interactive tool call never blocks silently.
list_members are masked by the API itself (e.g. u***@example.com) — this server never sees or handles unmasked member PII.get_site_settings by the API — there is no way to read them through this server.PATCH /api/v1/comments/{id}/moderate takes { "status": "approved" | "pending" | "spam" } per the API's own schema, so the tool's input is named to match.{ "data": ... }) — only the write endpoints (create_post, update_post, moderate_comment) wrap their response, matching the API's own inconsistency here (documented in openapi.yaml).upload_media tool — POST /api/v1/media takes multipart/form-data, which doesn't map cleanly onto typical MCP client transports. list_media is available for referencing media already uploaded through the dashboard. Contributions welcome if you need this.npm install
npm run typecheck # tsc --noEmit npm test # vitest — all tests run against a mocked fetch, no live API calls npm run build # emits dist/
The test suite covers: the Authorization: Bearer header on every request, each tool's exact method/URL/body mapping, error-code mapping for 401/403/429/500 responses, and the missing-env-var fail-fast path. No test in this repository makes a real network call.
MIT — see LICENSE.