Docstar

io.github.swayammaheshwari由 GitHub 账号 swayammaheshwari 发布v1.0.1更新于 Oct 29, 2026

Publish your API docs so AI assistants can look up endpoints and call your API correctly.

已验证Streamable HTTP可网页运行知识与记忆开发者工具

概览

AI 生成的概览

DocStar 的远程 MCP 服务器让助手搜索 DocStar 文档、浏览其 API 参考,并读取或修改已登录用户的文档集合。

可以做什么
  • 搜索 DocStar 文档与指南,并逐层浏览 API 参考
  • 列出组织与集合,查看集合的页面树和 MCP 服务器设置
  • 创建集合、页面和 API 端点文档,更新端点,批量发布页面
  • 以已登录用户身份用 GET、POST、PUT、PATCH 或 DELETE 调用任意 DocStar API 端点
  • 向 DocStar 团队发送缺陷报告、功能请求或文档反馈
  • 加载 quickstart 技能,获取查找和调用 DocStar 端点的操作说明
适用场景
适合在 DocStar 上维护 API 文档时使用:让助手查询 DocStar 的用法、查看或编辑集合与端点文档、发布页面,或调用没有专用工具的 DocStar 端点。
运行要求
远程 MCP 端点 mcp.docstar.io,使用 streamable HTTP,无需安装包或本地运行时。用户需已登录 DocStar,多数工具需要先用 list_organizations 取得组织 id。写入类工具需要集合的读写权限。
安装前请注意
这些工具作用于已登录用户的 DocStar 账号:可创建集合、页面和端点,更新端点,发布页面,并向 DocStar API 发送任意 POST、PUT、PATCH 或 DELETE 请求。发布会触发搜索索引并使内容公开;更新后端点会回到草稿状态,需重新发布。反馈内容会发送给 DocStar 团队。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Docstar,将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Web executable,通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

交给智能体安装

把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体,或 SourceWeft 对话中。智能体会读取该服务器的安装指南,先向你展示发布者、端点和登录方式,经你同意后再把它添加到你的客户端。密钥不会写进任何命令。

阅读 https://sourceweft.com/mcp/io.github.swayammaheshwari%2Fdocstar-mcp/install.md,并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前,先告诉我它的发布者、端点和登录方式,等我确认。不要把 API 密钥写进命令或我会提交的文件;问我密钥应从哪里获取。

查看智能体所依据的安装指南

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。

{
  "mcpServers": {
    "docstar": {
      "type": "http",
      "url": "https://mcp.docstar.io/mcp"
    }
  }
}

工具

15

search_docstar_docs

Search Docstar docs

Search DocStar's documentation and guides with a natural-language question, e.g. "how do I add a custom domain to my docs?" or "how does sign-in work on my MCP server?". Returns the most relevant passages with their page URLs. Use it for how-to and concept questions about DocStar. To find an API endpoint to call, use get_docstar_api_docs instead.

输入参数query必填collectionIds必填

get_docstar_api_docs

Get Docstar API docs

Browse the DocStar API reference one level at a time, when no dedicated tool on this server does what the user needs. Call it with {} for the top level: the modules (Collections, Pages, Endpoints, Published Versions, MCP Server, FAQs, Changelogs, Webhooks, URL Mappings, ...) with a short description and the endpoints they hold. Call it with `path` set to a module's link to see inside it, and with an endpoint's link ending in ".openapi.yml" to get that endpoint as an OpenAPI file: method, url, headers, query params, path variables, a body example and sample responses. Fetch only the level you need, and reuse what you already fetched.

输入参数path

call_any_api_of_docstar

Write to Docstar API

Send a POST, PUT, PATCH or DELETE request to any DocStar API endpoint for the signed-in user and return { status, ok, headers, data }. Use it for changes no dedicated tool covers, once you know the exact endpoint: from the quickstart-docstar skill's recipes, or looked up with get_docstar_api_docs. For GET requests use call_any_read_api_of_docstar. Never guess a URL. Pass the full url with path variables filled in, the method, the headers the endpoint's docs list, query params in `query` (not in the url) and the JSON body in `body`. Send the endpoint's auth header with an empty value "": the user's token is filled in automatically. Never put a token there yourself. Show the user what you're about to send and get a yes before any call that creates, changes, publishes or deletes something.

输入参数url必填bodyquerymethod必填headers必填

call_any_read_api_of_docstar

Read from Docstar API

Read from any DocStar API endpoint with a GET request for the signed-in user and return { status, ok, headers, data }. Use it for lookups no dedicated tool covers, once you know the exact endpoint: from the quickstart-docstar skill's recipes, or looked up with get_docstar_api_docs. Never guess a URL. Pass the full url with path variables filled in, the headers the endpoint's docs list, and query params in `query` (not in the url). Send the endpoint's auth header with an empty value "": the user's token is filled in automatically. Never put a token there yourself.

输入参数url必填queryheaders必填

list_organizations

List organizations

List the DocStar organizations (workspaces) the signed-in user belongs to, with each one's id and name (the list is at data.data). Start every DocStar task here: show the user the organizations and ask which one to use, unless there is only one. The chosen id is the orgId every other DocStar tool needs.

输入参数pageNoitemsPerPage

list_collections

List collections

List the collections (docs sites) in an organization, with each one's id, name, type and the caller's permission (1 read, 2 read-write). Each collection also has rootParentId: the parent pageId to use with create_page for a top-level page. Get orgId from list_organizations. Pass exclude=runCollection,environment to keep the response small. Ask the user which collection to work in.

输入参数orgId必填exclude

get_mcp_server

Get MCP server

Get a collection's MCP server: the url end users connect to, ready-to-paste snippets (a Claude Code command and a JSON config), callbackUrls for the OAuth app, the sign-in setup and any custom domain. Creates the server if it's missing and syncs its tools with the published endpoints. Call it after publishing to hand the user their MCP URL.

输入参数orgId必填collectionId必填

create_collection

Creates a new documentation collection in the organization together with its invisible root page (type 0), auto-assigns collection permissions to all org users based on their roles, and auto-provisions an MCP server for it in the background. The returned rootParentId is the page id under which top-level pages are created (via Create Page). SSO read-only organizations that already have a collection get 403. Slug is permanent; if omitted it is derived from the name, and a chosen slug that is taken returns 409. A body that fails validation returns 400 with the Joi message. Body fields: name (string, required) - Collection name, 3-55 characters. If the organization already has a collection with the same name (or a name like "name_N"), a numeric suffix is appended to the stored name (for example "Docs_2") instead of an error. slug (string, optional) - Permanent URL-safe identifier: lowercase letters, numbers and underscores only (no hyphens or spaces; 2-50 chars). Cannot be changed later. If omitted it is derived from the name (every run of other characters becomes an underscore), with a random suffix added when that slug is already taken. description (string, optional) - Collection description (null or empty allowed). isPublic (boolean, optional) - Accepted but ignored on create: a new collection is always created with isPublic true. Use Update Collection to change it. type (string, optional) - Collection type: 'page' (default, documentation pages), 'api' (API reference), 'blog', or 'embed'. customDomain (string, optional) - Custom domain to serve the public docs on, e.g. docs.example.com. Must be unique. path (string, optional) - Path prefix under the custom domain, e.g. 'api'. theme (string, optional) - Theme/brand color hex string, default '#f2994a'. favicon (string, optional) - Favicon URL. gtmId (string, optional) - Google Tag Manager ID. docProperties (object, optional) - Free-form do

输入参数nameslugtypeorgId必填description

create_endpoint

Creates a new API endpoint doc (page type 4, draft) as the last child of the given page. Requires READ_WRITE permission. Publish it afterwards with Update Endpoint State (parent page must be published). Body fields: name (string, optional) - Endpoint title shown in docs. urlName (string, optional) - URL slug; derived from name if omitted and made unique. requestType (string, optional) - HTTP method: GET, POST, PUT, PATCH, DELETE, etc. URL (string, optional) - Request URL (may contain {{variables}}). description (string, optional) - Endpoint description (HTML/text). body (object, optional) - Request body definition, e.g. {"type":"JSON","raw":{"rawType":"JSON","value":"{...}"}}; type can be none, JSON, TEXT, XML, HTML, multipart/form-data, application/x-www-form-urlencoded. headers (object, optional) - Headers keyed by name: {"Authorization":{"checked":"true","value":"Bearer {{token}}","description":""}}. params (object, optional) - Query params keyed by name, same shape as headers. pathVariables (object, optional) - Path variables keyed by index: {"0":{"checked":"notApplicable","key":"id","value":"","description":"Payment id"}}. bodyDescription (object, optional) - Per-field body docs: {"amount":{"value":100,"type":"number","description":"Amount in cents"}}. sampleResponse (array, optional) - Sample responses: [{"title":"Success","status":200,"description":"","data":{...}}]. authorizationData (object, optional) - Auth config object (or null). preScript (string, optional) - Pre-request JavaScript. postScript (string, optional) - Post-response test JavaScript. docViewData (array, optional) - Order of doc sections, e.g. [{&qu

输入参数URLbodynameorgId必填pageId必填paramsheadersdescriptionrequestTypepathVariablessampleResponse

update_endpoint

Partially updates an API endpoint doc; send only fields to change (same fields as Create Endpoint). Any update moves a non-pending endpoint back to draft (state 1), so republish with Update Endpoint State. Requires READ_WRITE permission. Only works for type 4/5 items (otherwise 400 "Wrong Route"). Invalid fields return 400 with the validation message. Body fields: name (string, optional) - Endpoint title shown in docs. urlName (string, optional) - New URL slug; made unique among siblings. Left unchanged if omitted (it is not re-derived from name). requestType (string, optional) - HTTP method: GET, POST, PUT, PATCH, DELETE, etc. URL (string, optional) - Request URL (may contain {{variables}}). description (string, optional) - Endpoint description (HTML/text). body (object, optional) - Request body definition, e.g. {"type":"JSON","raw":{"rawType":"JSON","value":"{...}"}}; type can be none, JSON, TEXT, XML, HTML, multipart/form-data, application/x-www-form-urlencoded. headers (object, optional) - Headers keyed by name: {"Authorization":{"checked":"true","value":"Bearer {{token}}","description":""}}. params (object, optional) - Query params keyed by name, same shape as headers. pathVariables (object, optional) - Path variables keyed by index: {"0":{"checked":"notApplicable","key":"id","value":"","description":"Payment id"}}. bodyDescription (object, optional) - Per-field body docs: {"amount":{"value":100,"type":"number","description":"Amount in cents"}}. sampleResponse (array, optional) - Sample responses: [{"title":"Success","status":200,"description":"","data":{...}}]. authorizationData (object, optional) - Auth config object (or null). preScript (string, optional) - Pre-request JavaScript. postScript (string, optional) - Post-response test JavaScript. docViewData (array, optional) - Order of doc sections, e.g. [{"type":"host"},{"type":"body"},...]. state (number, optional) - 0 pending, 1 draft, 2 approved, 3 rejected. Accepted, but the server sets state 1 (draft) on

输入参数URLbodynameorgId必填paramsheadersendpointId必填descriptionrequestTypepathVariablessampleResponse

create_page

Creates a new page as the last child of the page given in the path. Pass the collection's rootParentId (invisible root, from List Collections) to create a top-level page (type 1, parent page/module); pass any other page id to create a subpage (type 3). The type is decided by the parent, not the body. New pages start as draft (state 1); passing state 2 publishes immediately (parent must already be published) and returns publicUrl. Requires READ_WRITE permission. To write the page body afterwards use Update Page Content. Body fields: name (string, required) - Page title (min 1 char). urlName (string, optional) - URL slug (min 3 chars); made unique among siblings. Derived from name if omitted. contents (string, optional) - Legacy initial HTML/text contents column (editor content is stored in doc-rtc; prefer Update Page Content). state (number, optional) - 0 pending, 1 draft (default), 2 approved/published (publishes right away; parent must be published), 3 rejected. meta (object, optional) - Page metadata object (e.g. {"description":"...","tags":["setup"]}); may include editable boolean. sharingMeta (object, optional) - Sharing settings object, e.g. {"editable":true}. pageType (number, optional) - Accepted (1 or 3) but ignored; the type is derived from the parent page. uniqueTabId (string, optional) - Optional browser tab id used only to suppress the realtime socket echo to the calling tab; agents can omit it. POST https://api.docstar.io/orgs/:orgId/pages/:pageId

输入参数nameorgId必填statepageId必填urlNamecontents

publish_pages

Publishes many pages/endpoints at once: each page in pageIds gets a new default published version and state 2 (approved/published). rootParentId (the collection's invisible root page) must also be supplied so it is marked published. Requires READ_WRITE permission. Triggers background search indexing, translation and AI SEO/meta generation, and clears the docs cache. Get ids from List Collections (rootParentId) and List Collection Pages. Body fields: rootParentId (string, required) - 12-char id of the collection's invisible root page (type 0). pageIds (array<string>, required) - Non-empty array of 12-char page/endpoint ids to publish. PATCH https://api.docstar.io/orgs/:orgId/collections/:collectionId/bulkPublish

输入参数orgId必填pageIdscollectionId必填rootParentId

list_collection_pages

Returns all non-deleted pages of a collection (the full sidebar tree, flat) with id, name, urlName, type, parentId, child (ordered child ids), state, isPublished, meta, sharingMeta and the endpoint requestType for endpoint pages. Page types: 0 invisible root, 1 parent page/module (top-level), 3 subpage, 4 API endpoint, 5 example/sample request of an endpoint. States: 0 pending, 1 draft, 2 approved/published, 3 rejected. Use it to find pageIds and the hierarchy. Collection must belong to the caller's org. GET https://api.docstar.io/orgs/:orgId/collection/:collectionId/pages

输入参数orgId必填collectionId必填

send_feedback_to_docstar

Sends feedback straight to the DocStar team. Use it when the user wants to report a bug, request a feature, flag wrong or missing docs, or share praise about DocStar or this MCP server. Write the message in the user's words and add any useful context, such as the tool or endpoint involved and what went wrong. Who sent it (user and organization) is taken from the signed-in user's token, never from the body. Returns 200 once delivered; 400 for an invalid body; 502 if it couldn't be delivered, in which case tell the user and try again later. Authenticate with the proxy_auth_token header (an `Authorization: Bearer hm_...` DocStar API token or a DocStar JWT is also accepted). Body fields: message (string, required) - The feedback, 1 to 5000 characters.type (string, optional) - One of bug, feature_request, docs, praise, other; default other.collectionId (string, optional) - The collection the feedback is about, if any.context (object, optional) - Extra details, e.g. the tool or endpoint used, the request and the error. POST https://api.docstar.io/feedback/ai

输入参数typecontextmessagecollectionId

quickstart-docstar

Quickstart Docstar

Quickstart for the Docstar API: find any endpoint in the Docstar docs and call it for the signed-in user. Load this before any Docstar task. Call this to load the full instructions for this skill before doing the task.

无需输入参数。

版本历史

2
  1. v1.0.1最新Oct 29, 2026
  2. v1.0.0Oct 26, 2026

评分

已满足适用检查项的 83%

该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。

  • 近 30 天内可连通必需25 分
  • 已声明认证方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 图标10 分
  • 文档15 分

    请声明包含 README 的 GitHub 仓库,或在 OAuth 受保护资源元数据中提供 resource_documentation。

  • 许可证不适用

    托管服务按其提供方的条款使用,因此这里不要求许可证。此检查项不计入百分比。