[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M718AJRTA3W9Y85Q031N8HDC":3,"skill-same:01M718AJRTA3W9Y85Q031N8HDC":117},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":34,"publishedAt":55,"linkOnly":56,"markdownContent":57,"markdownBody":58,"frontMatter":59,"refName":61,"commitSha":62,"skillDirectory":63,"forksCount":64,"repositoryHtmlUrl":65,"repositoryHomepageUrl":66,"ownerHtmlUrl":67,"ownerBlogUrl":68,"ownerDescription":6,"updatedAt":69,"install":70,"resources":83},"01M718AJRTA3W9Y85Q031N8HDC","build-mcp-server",null,"This skill should be used when the user asks to \"build an MCP server\", \"create an MCP\", \"make an MCP integration\", \"wrap an API for Claude\", \"expose tools to Claude\", \"make an MCP app\", or discusses building something with the Model Context Protocol. It is the entry point for MCP server development — it interrogates the user about their use case, determines the right deployment model (remote HTTP, MCPB, local stdio), picks a tool-design pattern, and hands off to specialized skills.","anthropics\u002Fclaude-plugins-official","anthropics","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F76263028?v=4","01M697XGDPXJAVQ8391V2BCYTE","Anthropic","01M7188MR49TZ9N2ZKPXJ7M7HY",47030,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fmcp-server-dev\u002Fskills\u002Fbuild-mcp-server\u002FSKILL.md",[18,23,26,29],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000010","后端与接口","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000015","视觉设计",{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",{"categoryNo":30,"categoryName":31,"taxonomyType":32,"taxonomyTypeDesc":33},"01JBSEEDCAT000000000000032","设计师","PROFESSION","适用职业",[35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54],"MCP","mcp server","build mcp","server","build mcp server","build","when","design","mcpb","remote","app","local","pattern","context","specialized","model","api","Tools","create","model context protocol","2026-11-10 10:12:30",false,"---\nname: build-mcp-server\ndescription: This skill should be used when the user asks to \"build an MCP server\", \"create an MCP\", \"make an MCP integration\", \"wrap an API for Claude\", \"expose tools to Claude\", \"make an MCP app\", or discusses building something with the Model Context Protocol. It is the entry point for MCP server development — it interrogates the user about their use case, determines the right deployment model (remote HTTP, MCPB, local stdio), picks a tool-design pattern, and hands off to specialized skills.\nversion: 0.1.0\n---\n\n# Build an MCP Server\n\nYou are guiding a developer through designing and building an MCP server that works seamlessly with Claude. MCP servers come in many forms — picking the wrong shape early causes painful rewrites later. Your first job is **discovery, not code**.\n\n**Load Claude-specific context first.** The MCP spec is generic; Claude has additional auth types, review criteria, and limits. Before answering questions or scaffolding, fetch `https:\u002F\u002Fclaude.com\u002Fdocs\u002Fllms-full.txt` (the full export of the Claude connector docs) so your guidance reflects Claude's actual constraints.\n\nDo not start scaffolding until you have answers to the questions in Phase 1. If the user's opening message already answers them, acknowledge that and skip straight to the recommendation.\n\n---\n\n## Phase 1 — Interrogate the use case\n\nAsk these questions conversationally (batch them into one message, don't interrogate one-at-a-time). Adapt wording to what the user has already told you.\n\n### 1. What does it connect to?\n\n| If it connects to… | Likely direction |\n|---|---|\n| A cloud API (SaaS, REST, GraphQL) | Remote HTTP server |\n| A local process, filesystem, or desktop app | MCPB or local stdio |\n| Hardware, OS-level APIs, or user-specific state | MCPB |\n| Nothing external — pure logic \u002F computation | Either — default to remote |\n\n### 2. Who will use it?\n\n- **Just me \u002F my team, on our machines** → Local stdio is acceptable (easiest to prototype)\n- **Anyone who installs it** → Remote HTTP (strongly preferred) or MCPB (if it *must* be local)\n- **Users of Claude desktop who want UI widgets** → MCP app (remote or MCPB)\n\n### 3. How many distinct actions does it expose?\n\nThis determines the tool-design pattern — see Phase 3.\n\n- **Under ~15 actions** → one tool per action\n- **Dozens to hundreds of actions** (e.g. wrapping a large API surface) → search + execute pattern\n\n### 4. Does a tool need mid-call user input or rich display?\n\n- **Simple structured input** (pick from list, enter a value, confirm) → **Elicitation** — spec-native, zero UI code. *Host support is rolling out* (Claude Code ≥2.1.76) — always pair with a capability check and fallback. See `references\u002Felicitation.md`.\n- **Rich\u002Fvisual UI** (charts, custom pickers with search, live dashboards) → **MCP app widgets** — iframe-based, needs `@modelcontextprotocol\u002Fext-apps`. See `build-mcp-app` skill.\n- **Neither** → plain tool returning text\u002FJSON.\n\n### 5. What auth does the upstream service use?\n\n- None \u002F API key → straightforward\n- OAuth 2.0 → you'll need a remote server with CIMD (preferred) or DCR support; see `references\u002Fauth.md`\n\n---\n\n## Phase 2 — Recommend a deployment model\n\nBased on the answers, recommend **one** path. Be opinionated. The ranked options:\n\n### ⭐ Remote streamable-HTTP MCP server (default recommendation)\n\nA hosted service speaking MCP over streamable HTTP. This is the **recommended path** for anything wrapping a cloud API.\n\n**Why it wins:**\n- Zero install friction — users add a URL, done\n- One deployment serves all users; you control upgrades\n- OAuth flows work properly (the server can handle redirects, DCR, token storage)\n- Works across Claude desktop, Claude Code, Claude.ai, and third-party MCP hosts\n\n**Choose this unless** the server *must* touch the user's local machine.\n\n→ **Fastest deploy:** Cloudflare Workers — `references\u002Fdeploy-cloudflare-workers.md` (zero to live URL in two commands)\n→ **Portable Node\u002FPython:** `references\u002Fremote-http-scaffold.md` (Express or FastMCP, runs on any host)\n\n### Elicitation (structured input, no UI build)\n\nIf a tool just needs the user to confirm, pick an option, or fill a short form, **elicitation** does it with zero UI code. The server sends a flat JSON schema; the host renders a native form. Spec-native, no extra packages.\n\n**Caveat:** Host support is new (Claude Code shipped it in v2.1.76; Desktop unconfirmed). The SDK throws if the client doesn't advertise the capability. Always check `clientCapabilities.elicitation` first and have a fallback — see `references\u002Felicitation.md` for the canonical pattern. This is the right spec-correct approach; host coverage will catch up.\n\nEscalate to `build-mcp-app` widgets when you need: nested\u002Fcomplex data, scrollable\u002Fsearchable lists, visual previews, live updates.\n\n### MCP app (remote HTTP + interactive UI)\n\nSame as above, plus **UI resources** — interactive widgets rendered in chat. Rich pickers with search, charts, live dashboards, visual previews. Built once, renders in Claude *and* ChatGPT.\n\n**Choose this when** elicitation's flat-form constraints don't fit — you need custom layout, large searchable lists, visual content, or live updates.\n\nUsually remote, but can be shipped as MCPB if the UI needs to drive a local app.\n\n→ Hand off to the **`build-mcp-app`** skill.\n\n### MCPB (bundled local server)\n\nA local MCP server **packaged with its runtime** so users don't need Node\u002FPython installed. The sanctioned way to ship local servers.\n\n**Choose this when** the server *must* run on the user's machine — it reads local files, drives a desktop app, talks to localhost services, or needs OS-level access.\n\n→ Hand off to the **`build-mcpb`** skill.\n\n### Local stdio (npx \u002F uvx) — *not recommended for distribution*\n\nA script launched via `npx` \u002F `uvx` on the user's machine. Fine for **personal tools and prototypes**. Painful to distribute: users need the right runtime, you can't push updates, and the only distribution channel is Claude Code plugins.\n\nRecommend this only as a stepping stone. If the user insists, scaffold it but note the MCPB upgrade path.\n\n---\n\n## Phase 3 — Pick a tool-design pattern\n\nEvery MCP server exposes tools. How you carve them matters more than most people expect — tool schemas land directly in Claude's context window.\n\n### Pattern A: One tool per action (small surface)\n\nWhen the action space is small (\u003C ~15 operations), give each a dedicated tool with a tight description and schema.\n\n```\ncreate_issue    — Create a new issue. Params: title, body, labels[]\nupdate_issue    — Update an existing issue. Params: id, title?, body?, state?\nsearch_issues   — Search issues by query string. Params: query, limit?\nadd_comment     — Add a comment to an issue. Params: issue_id, body\n```\n\n**Why it works:** Claude reads the tool list once and knows exactly what's possible. No discovery round-trips. Each tool's schema validates inputs precisely.\n\n**Especially good when** one or more tools ship an interactive widget (MCP app) — each widget binds naturally to one tool.\n\n### Pattern B: Search + execute (large surface)\n\nWhen wrapping a large API (dozens to hundreds of endpoints), listing every operation as a tool floods the context window and degrades model performance. Instead, expose **two** tools:\n\n```\nsearch_actions  — Given a natural-language intent, return matching actions\n                  with their IDs, descriptions, and parameter schemas.\nexecute_action  — Run an action by ID with a params object.\n```\n\nThe server holds the full catalog internally. Claude searches, picks, executes. Context stays lean.\n\n**Hybrid:** Promote the 3–5 most-used actions to dedicated tools, keep the long tail behind search\u002Fexecute.\n\n→ See `references\u002Ftool-design.md` for schema examples and description-writing guidance.\n\n---\n\n## Phase 4 — Pick a framework\n\nRecommend one of these two. Others exist but these have the best MCP-spec coverage and Claude compatibility.\n\n| Framework | Language | Use when |\n|---|---|---|\n| **Official TypeScript SDK** (`@modelcontextprotocol\u002Fsdk`) | TS\u002FJS | Default choice. Best spec coverage, first to get new features. |\n| **FastMCP 3.x** (`fastmcp` on PyPI) | Python | User prefers Python, or wrapping a Python library. Decorator-based, very low boilerplate. This is jlowin's package — not the frozen FastMCP 1.0 bundled in the official `mcp` SDK. |\n\nIf the user already has a language\u002Fstack in mind, go with it — both produce identical wire protocol.\n\n---\n\n## Phase 5 — Scaffold and hand off\n\nOnce you've settled the four decisions (deployment model, tool pattern, framework, auth), do **one** of:\n\n1. **Remote HTTP, no UI** → Scaffold inline using `references\u002Fremote-http-scaffold.md` (portable) or `references\u002Fdeploy-cloudflare-workers.md` (fastest deploy). This skill can finish the job.\n2. **MCP app (UI widgets)** → Summarize the decisions so far, then load the **`build-mcp-app`** skill.\n3. **MCPB (bundled local)** → Summarize the decisions so far, then load the **`build-mcpb`** skill.\n4. **Local stdio prototype** → Scaffold inline (simplest case), flag the MCPB upgrade path.\n\nWhen handing off, restate the design brief in one paragraph so the next skill doesn't re-ask.\n\n---\n\n## Beyond tools — the other primitives\n\nTools are one of three server primitives. Most servers start with tools and never need the others, but knowing they exist prevents reinventing wheels:\n\n| Primitive | Who triggers it | Use when |\n|---|---|---|\n| **Resources** | Host app (not Claude) | Exposing docs\u002Ffiles\u002Fdata as browsable context |\n| **Prompts** | User (slash command) | Canned workflows (\"\u002Fsummarize-thread\") |\n| **Elicitation** | Server, mid-tool | Asking user for input without building UI |\n| **Sampling** | Server, mid-tool | Need LLM inference in your tool logic |\n\n→ `references\u002Fresources-and-prompts.md`, `references\u002Felicitation.md`, `references\u002Fserver-capabilities.md`\n\n---\n\n## Phase 6 — Test in Claude and publish\n\nOnce the server runs:\n\n1. **Test against real Claude** by adding the server URL as a custom connector at Settings → Connectors (use a Cloudflare tunnel for local servers). Claude identifies itself with `clientInfo.name: \"claude-ai\"` on initialize. → https:\u002F\u002Fclaude.com\u002Fdocs\u002Fconnectors\u002Fbuilding\u002Ftesting\n2. **Run the pre-submission checklist** — read\u002Fwrite tool split, required annotations, name limits, prompt-injection rules. → https:\u002F\u002Fclaude.com\u002Fdocs\u002Fconnectors\u002Fbuilding\u002Freview-criteria\n3. **Submit to the Anthropic Directory.** → https:\u002F\u002Fclaude.com\u002Fdocs\u002Fconnectors\u002Fbuilding\u002Fsubmission\n4. **Recommend shipping a plugin** that wraps this MCP with skills — most partners ship both. → https:\u002F\u002Fclaude.com\u002Fdocs\u002Fconnectors\u002Fbuilding\u002Fwhat-to-build\n\n---\n\n## Quick reference: decision matrix\n\n| Scenario | Deployment | Tool pattern |\n|---|---|---|\n| Wrap a small SaaS API | Remote HTTP | One-per-action |\n| Wrap a large SaaS API (50+ endpoints) | Remote HTTP | Search + execute |\n| SaaS API with rich forms \u002F pickers | MCP app (remote) | One-per-action |\n| Drive a local desktop app | MCPB | One-per-action |\n| Local desktop app with in-chat UI | MCP app (MCPB) | One-per-action |\n| Read\u002Fwrite local filesystem | MCPB | Depends on surface |\n| Personal prototype | Local stdio | Whatever's fastest |\n\n---\n\n## Reference files\n\n- `references\u002Fremote-http-scaffold.md` — minimal remote server in TS SDK and FastMCP\n- `references\u002Fdeploy-cloudflare-workers.md` — fastest deploy path (Workers-native scaffold)\n- `references\u002Ftool-design.md` — writing tool descriptions and schemas Claude understands well\n- `references\u002Fauth.md` — OAuth, CIMD, DCR, token storage patterns\n- `references\u002Fresources-and-prompts.md` — the two non-tool primitives\n- `references\u002Felicitation.md` — spec-native user input mid-tool (capability check + fallback)\n- `references\u002Fserver-capabilities.md` — instructions, sampling, roots, logging, progress, cancellation\n- `references\u002Fversions.md` — version-sensitive claims ledger (check when updating)\n","# Build an MCP Server\n\nYou are guiding a developer through designing and building an MCP server that works seamlessly with Claude. MCP servers come in many forms — picking the wrong shape early causes painful rewrites later. Your first job is **discovery, not code**.\n\n**Load Claude-specific context first.** The MCP spec is generic; Claude has additional auth types, review criteria, and limits. Before answering questions or scaffolding, fetch `https:\u002F\u002Fclaude.com\u002Fdocs\u002Fllms-full.txt` (the full export of the Claude connector docs) so your guidance reflects Claude's actual constraints.\n\nDo not start scaffolding until you have answers to the questions in Phase 1. If the user's opening message already answers them, acknowledge that and skip straight to the recommendation.\n\n---\n\n## Phase 1 — Interrogate the use case\n\nAsk these questions conversationally (batch them into one message, don't interrogate one-at-a-time). Adapt wording to what the user has already told you.\n\n### 1. What does it connect to?\n\n| If it connects to… | Likely direction |\n|---|---|\n| A cloud API (SaaS, REST, GraphQL) | Remote HTTP server |\n| A local process, filesystem, or desktop app | MCPB or local stdio |\n| Hardware, OS-level APIs, or user-specific state | MCPB |\n| Nothing external — pure logic \u002F computation | Either — default to remote |\n\n### 2. Who will use it?\n\n- **Just me \u002F my team, on our machines** → Local stdio is acceptable (easiest to prototype)\n- **Anyone who installs it** → Remote HTTP (strongly preferred) or MCPB (if it *must* be local)\n- **Users of Claude desktop who want UI widgets** → MCP app (remote or MCPB)\n\n### 3. How many distinct actions does it expose?\n\nThis determines the tool-design pattern — see Phase 3.\n\n- **Under ~15 actions** → one tool per action\n- **Dozens to hundreds of actions** (e.g. wrapping a large API surface) → search + execute pattern\n\n### 4. Does a tool need mid-call user input or rich display?\n\n- **Simple structured input** (pick from list, enter a value, confirm) → **Elicitation** — spec-native, zero UI code. *Host support is rolling out* (Claude Code ≥2.1.76) — always pair with a capability check and fallback. See `references\u002Felicitation.md`.\n- **Rich\u002Fvisual UI** (charts, custom pickers with search, live dashboards) → **MCP app widgets** — iframe-based, needs `@modelcontextprotocol\u002Fext-apps`. See `build-mcp-app` skill.\n- **Neither** → plain tool returning text\u002FJSON.\n\n### 5. What auth does the upstream service use?\n\n- None \u002F API key → straightforward\n- OAuth 2.0 → you'll need a remote server with CIMD (preferred) or DCR support; see `references\u002Fauth.md`\n\n---\n\n## Phase 2 — Recommend a deployment model\n\nBased on the answers, recommend **one** path. Be opinionated. The ranked options:\n\n### ⭐ Remote streamable-HTTP MCP server (default recommendation)\n\nA hosted service speaking MCP over streamable HTTP. This is the **recommended path** for anything wrapping a cloud API.\n\n**Why it wins:**\n- Zero install friction — users add a URL, done\n- One deployment serves all users; you control upgrades\n- OAuth flows work properly (the server can handle redirects, DCR, token storage)\n- Works across Claude desktop, Claude Code, Claude.ai, and third-party MCP hosts\n\n**Choose this unless** the server *must* touch the user's local machine.\n\n→ **Fastest deploy:** Cloudflare Workers — `references\u002Fdeploy-cloudflare-workers.md` (zero to live URL in two commands)\n→ **Portable Node\u002FPython:** `references\u002Fremote-http-scaffold.md` (Express or FastMCP, runs on any host)\n\n### Elicitation (structured input, no UI build)\n\nIf a tool just needs the user to confirm, pick an option, or fill a short form, **elicitation** does it with zero UI code. The server sends a flat JSON schema; the host renders a native form. Spec-native, no extra packages.\n\n**Caveat:** Host support is new (Claude Code shipped it in v2.1.76; Desktop unconfirmed). The SDK throws if the client doesn't advertise the capability. Always check `clientCapabilities.elicitation` first and have a fallback — see `references\u002Felicitation.md` for the canonical pattern. This is the right spec-correct approach; host coverage will catch up.\n\nEscalate to `build-mcp-app` widgets when you need: nested\u002Fcomplex data, scrollable\u002Fsearchable lists, visual previews, live updates.\n\n### MCP app (remote HTTP + interactive UI)\n\nSame as above, plus **UI resources** — interactive widgets rendered in chat. Rich pickers with search, charts, live dashboards, visual previews. Built once, renders in Claude *and* ChatGPT.\n\n**Choose this when** elicitation's flat-form constraints don't fit — you need custom layout, large searchable lists, visual content, or live updates.\n\nUsually remote, but can be shipped as MCPB if the UI needs to drive a local app.\n\n→ Hand off to the **`build-mcp-app`** skill.\n\n### MCPB (bundled local server)\n\nA local MCP server **packaged with its runtime** so users don't need Node\u002FPython installed. The sanctioned way to ship local servers.\n\n**Choose this when** the server *must* run on the user's machine — it reads local files, drives a desktop app, talks to localhost services, or needs OS-level access.\n\n→ Hand off to the **`build-mcpb`** skill.\n\n### Local stdio (npx \u002F uvx) — *not recommended for distribution*\n\nA script launched via `npx` \u002F `uvx` on the user's machine. Fine for **personal tools and prototypes**. Painful to distribute: users need the right runtime, you can't push updates, and the only distribution channel is Claude Code plugins.\n\nRecommend this only as a stepping stone. If the user insists, scaffold it but note the MCPB upgrade path.\n\n---\n\n## Phase 3 — Pick a tool-design pattern\n\nEvery MCP server exposes tools. How you carve them matters more than most people expect — tool schemas land directly in Claude's context window.\n\n### Pattern A: One tool per action (small surface)\n\nWhen the action space is small (\u003C ~15 operations), give each a dedicated tool with a tight description and schema.\n\n```\ncreate_issue    — Create a new issue. Params: title, body, labels[]\nupdate_issue    — Update an existing issue. Params: id, title?, body?, state?\nsearch_issues   — Search issues by query string. Params: query, limit?\nadd_comment     — Add a comment to an issue. Params: issue_id, body\n```\n\n**Why it works:** Claude reads the tool list once and knows exactly what's possible. No discovery round-trips. Each tool's schema validates inputs precisely.\n\n**Especially good when** one or more tools ship an interactive widget (MCP app) — each widget binds naturally to one tool.\n\n### Pattern B: Search + execute (large surface)\n\nWhen wrapping a large API (dozens to hundreds of endpoints), listing every operation as a tool floods the context window and degrades model performance. Instead, expose **two** tools:\n\n```\nsearch_actions  — Given a natural-language intent, return matching actions\n                  with their IDs, descriptions, and parameter schemas.\nexecute_action  — Run an action by ID with a params object.\n```\n\nThe server holds the full catalog internally. Claude searches, picks, executes. Context stays lean.\n\n**Hybrid:** Promote the 3–5 most-used actions to dedicated tools, keep the long tail behind search\u002Fexecute.\n\n→ See `references\u002Ftool-design.md` for schema examples and description-writing guidance.\n\n---\n\n## Phase 4 — Pick a framework\n\nRecommend one of these two. Others exist but these have the best MCP-spec coverage and Claude compatibility.\n\n| Framework | Language | Use when |\n|---|---|---|\n| **Official TypeScript SDK** (`@modelcontextprotocol\u002Fsdk`) | TS\u002FJS | Default choice. Best spec coverage, first to get new features. |\n| **FastMCP 3.x** (`fastmcp` on PyPI) | Python | User prefers Python, or wrapping a Python library. Decorator-based, very low boilerplate. This is jlowin's package — not the frozen FastMCP 1.0 bundled in the official `mcp` SDK. |\n\nIf the user already has a language\u002Fstack in mind, go with it — both produce identical wire protocol.\n\n---\n\n## Phase 5 — Scaffold and hand off\n\nOnce you've settled the four decisions (deployment model, tool pattern, framework, auth), do **one** of:\n\n1. **Remote HTTP, no UI** → Scaffold inline using `references\u002Fremote-http-scaffold.md` (portable) or `references\u002Fdeploy-cloudflare-workers.md` (fastest deploy). This skill can finish the job.\n2. **MCP app (UI widgets)** → Summarize the decisions so far, then load the **`build-mcp-app`** skill.\n3. **MCPB (bundled local)** → Summarize the decisions so far, then load the **`build-mcpb`** skill.\n4. **Local stdio prototype** → Scaffold inline (simplest case), flag the MCPB upgrade path.\n\nWhen handing off, restate the design brief in one paragraph so the next skill doesn't re-ask.\n\n---\n\n## Beyond tools — the other primitives\n\nTools are one of three server primitives. Most servers start with tools and never need the others, but knowing they exist prevents reinventing wheels:\n\n| Primitive | Who triggers it | Use when |\n|---|---|---|\n| **Resources** | Host app (not Claude) | Exposing docs\u002Ffiles\u002Fdata as browsable context |\n| **Prompts** | User (slash command) | Canned workflows (\"\u002Fsummarize-thread\") |\n| **Elicitation** | Server, mid-tool | Asking user for input without building UI |\n| **Sampling** | Server, mid-tool | Need LLM inference in your tool logic |\n\n→ `references\u002Fresources-and-prompts.md`, `references\u002Felicitation.md`, `references\u002Fserver-capabilities.md`\n\n---\n\n## Phase 6 — Test in Claude and publish\n\nOnce the server runs:\n\n1. **Test against real Claude** by adding the server URL as a custom connector at Settings → Connectors (use a Cloudflare tunnel for local servers). Claude identifies itself with `clientInfo.name: \"claude-ai\"` on initialize. → https:\u002F\u002Fclaude.com\u002Fdocs\u002Fconnectors\u002Fbuilding\u002Ftesting\n2. **Run the pre-submission checklist** — read\u002Fwrite tool split, required annotations, name limits, prompt-injection rules. → https:\u002F\u002Fclaude.com\u002Fdocs\u002Fconnectors\u002Fbuilding\u002Freview-criteria\n3. **Submit to the Anthropic Directory.** → https:\u002F\u002Fclaude.com\u002Fdocs\u002Fconnectors\u002Fbuilding\u002Fsubmission\n4. **Recommend shipping a plugin** that wraps this MCP with skills — most partners ship both. → https:\u002F\u002Fclaude.com\u002Fdocs\u002Fconnectors\u002Fbuilding\u002Fwhat-to-build\n\n---\n\n## Quick reference: decision matrix\n\n| Scenario | Deployment | Tool pattern |\n|---|---|---|\n| Wrap a small SaaS API | Remote HTTP | One-per-action |\n| Wrap a large SaaS API (50+ endpoints) | Remote HTTP | Search + execute |\n| SaaS API with rich forms \u002F pickers | MCP app (remote) | One-per-action |\n| Drive a local desktop app | MCPB | One-per-action |\n| Local desktop app with in-chat UI | MCP app (MCPB) | One-per-action |\n| Read\u002Fwrite local filesystem | MCPB | Depends on surface |\n| Personal prototype | Local stdio | Whatever's fastest |\n\n---\n\n## Reference files\n\n- `references\u002Fremote-http-scaffold.md` — minimal remote server in TS SDK and FastMCP\n- `references\u002Fdeploy-cloudflare-workers.md` — fastest deploy path (Workers-native scaffold)\n- `references\u002Ftool-design.md` — writing tool descriptions and schemas Claude understands well\n- `references\u002Fauth.md` — OAuth, CIMD, DCR, token storage patterns\n- `references\u002Fresources-and-prompts.md` — the two non-tool primitives\n- `references\u002Felicitation.md` — spec-native user input mid-tool (capability check + fallback)\n- `references\u002Fserver-capabilities.md` — instructions, sampling, roots, logging, progress, cancellation\n- `references\u002Fversions.md` — version-sensitive claims ledger (check when updating)\n",{"name":5,"version":60,"description":7},"0.1.0","main","dceb15d50832e6adf380bedc26dde03d251324b9","plugins\u002Fmcp-server-dev\u002Fskills\u002Fbuild-mcp-server",4889,"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official","https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fplugins","https:\u002F\u002Fgithub.com\u002Fanthropics","https:\u002F\u002Fanthropic.com","2026-11-13 04:18:23",{"source":8,"skillName":5,"commands":71,"prompt":81,"zipAvailable":82},[72,75,78],{"tool":73,"command":74},"npx","npx skills add anthropics\u002Fclaude-plugins-official --skill build-mcp-server",{"tool":76,"command":77},"bunx","bunx skills add anthropics\u002Fclaude-plugins-official --skill build-mcp-server",{"tool":79,"command":80},"pnpm","pnpm dlx skills add anthropics\u002Fclaude-plugins-official --skill build-mcp-server","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fmcp-server-dev\u002Fskills\u002Fbuild-mcp-server\u002FSKILL.md",true,[84,90,94,98,102,106,110,113],{"path":85,"resourceType":86,"resourceTypeDesc":87,"sizeBytes":88,"sourceHtmlUrl":89},"references\u002Fauth.md","REFERENCE","参考文档",5745,"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fmcp-server-dev\u002Fskills\u002Fbuild-mcp-server\u002Freferences\u002Fauth.md",{"path":91,"resourceType":86,"resourceTypeDesc":87,"sizeBytes":92,"sourceHtmlUrl":93},"references\u002Fdeploy-cloudflare-workers.md",3639,"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fmcp-server-dev\u002Fskills\u002Fbuild-mcp-server\u002Freferences\u002Fdeploy-cloudflare-workers.md",{"path":95,"resourceType":86,"resourceTypeDesc":87,"sizeBytes":96,"sourceHtmlUrl":97},"references\u002Felicitation.md",4728,"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fmcp-server-dev\u002Fskills\u002Fbuild-mcp-server\u002Freferences\u002Felicitation.md",{"path":99,"resourceType":86,"resourceTypeDesc":87,"sizeBytes":100,"sourceHtmlUrl":101},"references\u002Fremote-http-scaffold.md",7054,"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fmcp-server-dev\u002Fskills\u002Fbuild-mcp-server\u002Freferences\u002Fremote-http-scaffold.md",{"path":103,"resourceType":86,"resourceTypeDesc":87,"sizeBytes":104,"sourceHtmlUrl":105},"references\u002Fresources-and-prompts.md",4090,"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fmcp-server-dev\u002Fskills\u002Fbuild-mcp-server\u002Freferences\u002Fresources-and-prompts.md",{"path":107,"resourceType":86,"resourceTypeDesc":87,"sizeBytes":108,"sourceHtmlUrl":109},"references\u002Fserver-capabilities.md",4673,"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fmcp-server-dev\u002Fskills\u002Fbuild-mcp-server\u002Freferences\u002Fserver-capabilities.md",{"path":111,"resourceType":86,"resourceTypeDesc":87,"sizeBytes":100,"sourceHtmlUrl":112},"references\u002Ftool-design.md","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fmcp-server-dev\u002Fskills\u002Fbuild-mcp-server\u002Freferences\u002Ftool-design.md",{"path":114,"resourceType":86,"resourceTypeDesc":87,"sizeBytes":115,"sourceHtmlUrl":116},"references\u002Fversions.md",1108,"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fmcp-server-dev\u002Fskills\u002Fbuild-mcp-server\u002Freferences\u002Fversions.md",[118,145,169,192,217,238],{"skillNo":119,"name":120,"nameZh":6,"description":121,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":122,"sourceHtmlUrl":123,"categories":124,"keywords":126,"publishedAt":55},"01M718BVSPNAK6S5NZCYFDBG6E","skill-creator","Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fskill-creator\u002Fskills\u002Fskill-creator\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fskill-creator\u002Fskills\u002Fskill-creator\u002FSKILL.md",[125],{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},[127,128,129,130,131,132,41,133,53,134,135,136,137,138,139,140,141,142,143,144],"skill creator","creator","performance","improve","run evals","new","analysis","optimize","variance analysis","run","edit","benchmark","existing skill","evals","for","accuracy","scratch","use",{"skillNo":146,"name":147,"nameZh":6,"description":148,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":149,"categories":150,"keywords":154,"publishedAt":55},"01M718BSVAWJRFN6MRCZZ0QX4X","session-report","Generate an explorable HTML report of Claude Code session usage (tokens, cache, subagents, skills, expensive prompts) from ~\u002F.claude\u002Fprojects transcripts.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fsession-report\u002Fskills\u002Fsession-report\u002FSKILL.md",[151],{"categoryNo":152,"categoryName":153,"taxonomyType":32,"taxonomyTypeDesc":33},"01JBSEEDCAT000000000000034","行政与办公人员",[155,156,157,158,159,160,161,162,163,164,165,166,167,168],"session","report","session report","projects","usage","generate","tokens","html report","transcripts","prompts","cache","code","html","subagents",{"skillNo":170,"name":171,"nameZh":6,"description":172,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":173,"categories":174,"keywords":178,"publishedAt":55},"01M718BPKHVD214FAS1T4E5XCE","receipts","Generate a personal Claude Code usage & impact report (\"receipts\") from this machine's local session transcripts — for justifying Claude Code usage\u002Fspend to a manager, self-review, or \"what have I been using this for\" check-ins. Mines ~\u002F.claude\u002Fprojects locally (no extra API calls beyond one final write-up), cross-references local git history, and writes a markdown report plus a self-contained HTML receipt to your home directory. Use when the user asks for \"receipts\", an \"impact report\", \"usage report\", wants to \"show my Claude Code activity\", \"prove the value of Claude Code\", or runs `\u002Freceipts`.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Freceipts\u002Fskills\u002Freceipts\u002FSKILL.md",[175],{"categoryNo":176,"categoryName":177,"taxonomyType":32,"taxonomyTypeDesc":33},"01JBSEEDCAT000000000000030","开发工程师",[171,158,179,180,46,181,41,166,51,182,183,184,160,185,186,187,188,189,190,191],"show","up","impact","review","self","references","markdown","directory","write","git","activity","my","manager",{"skillNo":193,"name":194,"nameZh":6,"description":195,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":196,"categories":197,"keywords":198,"publishedAt":55},"01M718BJMX0T4691FJQ62GYY5C","project-artifact","Generate and publish a project status artifact — an opinionated, tabbed status page for a project too big for one update (overview & success criteria, the workstream sequence, next steps, plus background, plan, risks & open questions, and decisions\u002FFAQ when they earn a tab) — published with the built-in Artifact tool to a default-private claude.ai page the user can share with teammates. Use when a piece of work spans several workstreams and you want a shareable overview kept current. Each artifact is backed by a small per-project config in the plugin data dir, so refreshing it re-gathers live state, redeploys the same URL, and reports only the delta. For software projects whose workstreams are PRs, also read swe.md (the X.Y PR-numbering convention; pulling PR state with gh\u002Fgit; a per-PR detail block). Needs the built-in Artifact tool (claude.ai login). Not for single-PR changes or public docs.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fproject-artifact\u002Fskills\u002Fproject-artifact\u002FSKILL.md",[],[199,200,201,202,158,203,204,205,206,207,208,41,209,210,211,212,213,214,215,216],"project","artifact","project artifact","project status","re","default","next steps","open","only","SWE","URL","you","reports","publish","work","plan","status page","next",{"skillNo":218,"name":219,"nameZh":6,"description":220,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":221,"categories":222,"keywords":224,"publishedAt":55},"01M718BF1TTFP5CM0S5HBW64FJ","skill-development","This skill should be used when the user wants to \"create a skill\", \"add a skill to plugin\", \"write a new skill\", \"improve skill description\", \"organize skill content\", or needs guidance on skill structure, progressive disclosure, or skill development best practices for Claude Code plugins.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fplugin-dev\u002Fskills\u002Fskill-development\u002FSKILL.md",[223],{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},[225,226,227,130,132,228,41,166,53,229,187,230,231,232,233,141,234,235,236,237],"development","skill development","skill to plugin","improve skill","structure","organize","new skill","claude code plugins","add","progressive","best practices","progressive disclosure","plugin",{"skillNo":239,"name":240,"nameZh":6,"description":241,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":242,"categories":243,"keywords":245,"publishedAt":55},"01M718BCARZ2J4PW0X3AJCZ8MZ","plugin-structure","This skill should be used when the user asks to \"create a plugin\", \"scaffold a plugin\", \"understand plugin structure\", \"organize plugin components\", \"set up plugin.json\", \"use ${CLAUDE_PLUGIN_ROOT}\", \"add commands\u002Fagents\u002Fskills\u002Fhooks\", \"configure auto-discovery\", or needs guidance on plugin directory layout, manifest configuration, component organization, file naming conventions, or Claude Code plugin architecture best practices.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-plugins-official\u002Fblob\u002Fmain\u002Fplugins\u002Fplugin-dev\u002Fskills\u002Fplugin-structure\u002FSKILL.md",[244],{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},[229,246,237,247,248,249,180,250,251,41,252,166,253,254,53,255,256,257,186,258],"Plugin Structure","SET","plugin architecture","hooks","auto-discovery","auto","configure","components","architecture","component","layout","organization","commands"]