原型 · 预览版:收录官方注册中心里的全部服务器,每 30 分钟同步一次。使用 Google 登录即可撰写评价。原型 · 每 30 分钟更新
PopNx MCP为使用 AI 的人准备的 MCP 服务器 · 用大白话讲清楚

StoryOS

✓ 服务方 · storyos.dev

作者:StoryFunnels版本 0.1.2i★ 8 (整个仓库的星标)i电脑或云端官方网站 ↗
注册名称dev.storyos/storyos

安装方法

What it does

Your StoryOS workspace (records, approvals and agents) for Claude, ChatGPT and any MCP client.

  • Connects 电脑或云端
  • Account or key Cloud: Sign in to the service when you connect (OAuth) · Local: Needs STORYOS_TOKEN
  • Category 其他

为什么排在这里

此项在“连接 AI 的应用”标签中,该标签不使用星标,按以下顺序排列

1. 在 Claude 连接器目录中否
2. 团队实测过否
3. 适合泰国用户否
4. 不计星标的使用量20/100热度 0 · 新鲜度 12 · 仍在响应 8 · 用户喜欢 +0
5. 登记日期2026-11-18

用户喜欢: 0 · 最多加 5 分(每人 1 分),不能用钱买。

质量: 基础检查通过 4/5 项(仅展示,不用于排序)

安装

选择你使用的 AI 应用,然后复制设置。

用一句话让 AI 帮你安装

请帮我安装 StoryOS。按照 https://mcp.popnx.com/i/storyos.md 操作,先把命令给我看,等我同意后再运行。

粘贴到能在你电脑上工作的 AI(如 Claude Code、Cursor)。它会读取我们的步骤,先给你看命令,等你同意。无需密钥,无需注册。 · 查看 AI 会读取的全文

通过 4/5 项检查 · 你的提问会经过开发者的服务器(mcp.storyos.dev) · ✓ 命令已检查 2026年11月19日

第一步:你用哪个 AI?

选择你用的应用,步骤会显示在下方。

可连接 Claude · 约 2 分钟

不需要密钥,不用安装程序 · 在 claude.ai 或桌面应用设置一次,即可在聊天中使用 · 免费账户可添加 1 个 · Team/Enterprise 方案由所有者(Owner)添加

  1. 打开 Claude 的 Connectors 页面

    CustomizeConnectors

    CustomizeConnectors1Add2https://…3Add4
  2. 点击“Add” → “Custom” → “Web”,填写这 2 项

    把这里的内容复制到 Claude 中同名的输入框,然后点击 Continue

    Name · 名称StoryOS
    Remote MCP server URL · 地址https://mcp.storyos.dev/mcp

    在 Claude 检测到的登录设置页点击 Continue · Authentication 按作者说明选择(Sign in now、Sign in when needed 或 No sign in),然后点击 Add

  3. 在聊天中开启

    +ConnectorsStoryOS

  4. 试一试

    用 StoryOS 告诉我你能帮我做什么,然后做一件给我看

    ✓ 如果 Claude 说正在使用 StoryOS,或回答中出现 StoryOS 的工具名称,就表示连接成功

菜单名称已于 2026年11月11日 对照 Anthropic 帮助页面核实 · 你的界面可能因版本略有不同

能否连接 ChatGPT 取决于方案

完整 MCP:网页版 ChatGPT Business、Enterprise 或 Edu · Pro:以读取/获取权限连接,修改数据的工具可能用不了 · Plus:官方文档尚未确认,如果在插件页看到添加自定义 MCP 的选项可以试试 · 仅限网页

  1. 在网页版 ChatGPT 打开添加 MCP 的页面

    Plugins+Add custom MCP server

    ChatGPTPlugins1+2https://…3Create4
  2. 填写名称和地址

    输入名称,在 Connection 下粘贴地址,勾选风险提示,然后点击 Create as a plugin

    Name · 名称StoryOS
    Connection · 地址https://mcp.storyos.dev/mcp
  3. 在聊天中选择 StoryOS

    开启新聊天,输入前先从插件列表中选择 StoryOS

  4. 试一试

    用 StoryOS 告诉我你能帮我做什么,然后做一件给我看

    ✓ 如果 ChatGPT 说正在使用 StoryOS,或回答中出现 StoryOS 的工具名称,就表示连接成功

菜单名称已于 2026年11月11日 对照 OpenAI 帮助页面核实 · 你的界面可能因版本略有不同

可连接 Claude Code · 约 2 分钟

不需要密钥,不用安装程序

  1. 打开终端

    Mac:按 Command+空格,输入 Terminal · Windows:打开 PowerShell

  2. 粘贴这条命令并按 Enter

    claude mcp add --transport http --scope user storyos https://mcp.storyos.dev/mcp
  3. 在 Claude Code 中输入 /mcp

    应能看到 StoryOS 在列表中并显示已连接

  4. 试一试

    用 StoryOS 告诉我你能帮我做什么,然后做一件给我看

    ✓ 如果 Claude Code 说正在使用 StoryOS,或回答中出现 StoryOS 的工具名称,就表示连接成功

菜单名称已于 2026年11月11日 对照 Anthropic 帮助页面核实 · 你的界面可能因版本略有不同

可连接 Cursor · 约 2 分钟

不需要密钥,不用安装程序

  1. 点击这个按钮打开 Cursor

    在 Cursor 中打开

    Cursor 会请你确认安装,点击 Install

  2. 在 Cursor 中以 Agent 模式打开 AI 聊天

  3. 试一试

    用 StoryOS 告诉我你能帮我做什么,然后做一件给我看

    ✓ 如果 Cursor 说正在使用 StoryOS,或回答中出现 StoryOS 的工具名称,就表示连接成功

菜单名称已于 2026年11月11日 对照 Cursor 帮助页面核实 · 你的界面可能因版本略有不同

可连接 VS Code · 约 2 分钟

不需要密钥,不用安装程序

  1. 点击这个按钮打开 VS Code

    在 VS Code 中打开

    VS Code 会请你确认安装,点击 Install

  2. 在 VS Code 中以 Agent 模式打开 AI 聊天

  3. 试一试

    用 StoryOS 告诉我你能帮我做什么,然后做一件给我看

    ✓ 如果 VS Code 说正在使用 StoryOS,或回答中出现 StoryOS 的工具名称,就表示连接成功

菜单名称已于 2026年11月11日 对照 VS Code 帮助页面核实 · 你的界面可能因版本略有不同

其他支持 MCP 的应用(如 Windsurf、Zed、Gemini CLI)添加方式各不相同,请在应用列表中查看

查看所有支持 MCP 的应用 →

通过网页连接的应用使用的地址

地址https://mcp.storyos.dev/mcp

JSON 设置在下方“开发者专用”中

卡住了?问问正在用 StoryOS 的人💬 讨论 →
开发者专用:终端命令和 JSON 设置

通过网页连接

Claude Code

claude mcp add --transport http --scope user storyos https://mcp.storyos.dev/mcp

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "storyos": {
      "url": "https://mcp.storyos.dev/mcp"
    }
  }
}

VS Code (mcp.json)

{
  "servers": {
    "storyos": {
      "type": "http",
      "url": "https://mcp.storyos.dev/mcp"
    }
  }
}

安装在电脑上

Claude Code

claude mcp add --scope user storyos --env STORYOS_TOKEN=your-value -- npx -y @storyos/mcp

Claude Desktop (claude_desktop_config.json)

{
  "mcpServers": {
    "storyos": {
      "command": "npx",
      "args": [
        "-y",
        "@storyos/mcp"
      ],
      "env": {
        "STORYOS_TOKEN": "your-value"
      }
    }
  }
}

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "storyos": {
      "command": "npx",
      "args": [
        "-y",
        "@storyos/mcp"
      ],
      "env": {
        "STORYOS_TOKEN": "your-value"
      }
    }
  }
}

VS Code (mcp.json)

{
  "servers": {
    "storyos": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@storyos/mcp"
      ],
      "env": {
        "STORYOS_TOKEN": "your-value"
      }
    }
  }
}

需要你填写的值

  • STORYOS_TOKEN 机密 必填 env
    StoryOS 个人 API 令牌(mn_pat_...),在 StoryOS 的 API tokens 下创建。仅限于一个工作区使用。
    查看英文原文StoryOS personal API token (mn_pat_...), created under API tokens in StoryOS. Scoped to one workspace.
  • STORYOS_URL env
    Base URL of the StoryOS API, for example https://app.storyos.dev for the cloud. When unset the server uses http://localhost:3001 (local development).
  • STORYOS_MCP_PROFILE env
    设置为 internal 以同时公开内部工具配置文件。保持未设置则使用默认工具。
    查看英文原文Set to internal to also expose the internal tool profile. Leave unset for the default tools.
  • WEB_URL env
    StoryOS 网页应用的基础 URL,用于在工具结果中构建链接。未设置时,服务器使用 http://localhost:3000
    查看英文原文Base URL of the StoryOS web app, used to build links in tool results. When unset the server uses http://localhost:3000.

说明为自动翻译,变量名保持原样;不确定时请查看原文。

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。

一览

GitHub 星标i
8 (整个仓库的星标)
该仓库包含多个 MCP 条目,星标在排名中只计算一次
每周下载量i
8/周
已验证域名
✓ storyos.dev
许可证i
未注明许可证
连接方式
电脑或云端
传输方式
stdio, streamable-http
最新版本i
0.1.2
分类i
其他
发布到注册中心i
2026年11月19日
代码最后更新i
2026年11月19日
最近检查
2026年11月19日
作者
StoryFunnels
软件包
npm: @storyos/mcp
注册状态
活跃

哪些应用可以使用它

依据这个服务器的连接方式和各应用自己的文档,检查于 2026年11月10日。

通过互联网连接,适用于

ClaudeChatGPTGooseLM StudioMsty StudioCherry StudioCursorVS Code + GitHub CopilotVisual StudioWindsurf (Devin Desktop)ZedKiroClineContinueJetBrains AI AssistantClaude CodeCodex CLIGemini CLIOpenCodenodecodeWarpLibreChatOpen WebUIRaycast

在你的电脑上运行,适用于

ClaudeGooseLM StudioMsty StudioCherry StudioCursorVS Code + GitHub CopilotVisual StudioWindsurf (Devin Desktop)ZedKiroClineContinueJetBrains AI AssistantClaude CodeCodex CLIGemini CLIOpenCodenodecodeWarpLibreChatRaycast

全部应用及其优缺点 →

它能做什么

服务器要求先登录,才会列出它的工具。

基础检查i

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20日

  • ✓ 通过

    作者与仓库一致

    注册名称已通过域名 storyos.dev 验证(注册中心会核实域名所有权)

  • ✓ 通过

    仍在维护

    代码最后更新于昨天

  • ! 注意

    有许可证

    GitHub 无法识别许可证类型;请打开文件自行查看

  • ✓ 通过

    端点有响应

    在 547 毫秒内响应(HTTP 401)

  • ✓ 通过

    使用 OAuth 登录

    声明支持 OAuth 登录,你的 AI 应用可以带你去登录

  • … 尚未检查

    工具描述中没有隐藏指令

    尚未检查;我们会在检查端点时读取工具列表

已公开的漏洞(OSV.dev)

OSV.dev 中没有 @storyos/mcp@0.1.2 的记录

只检查这个版本的包本身,不包括所有依赖 · 检查于 2026年11月19日

我们检查什么 → · JSON

作者:申请重新检查或报告信息错误
给作者:在 README 中加徽章

如果你是作者,可把这段代码贴到 README。徽章会链接回本页。

[![Featured on PopNx MCP](https://mcp.popnx.com/badge/storyos.svg)](https://mcp.popnx.com/en/servers/storyos)

讨论

提问、分享技巧或报告无法使用。按“有帮助”数量排序。

还没有帖子,欢迎第一个提问或分享。

此处内容由用户撰写,不代表 PopNx 团队。

登录后发帖

用户评价

还没有评价

登录后写评价

与“其他”中的其他服务器比较i

  • 星标数多于此分类中其他 20815 个服务器的 91%
  • 发布到注册中心的时间比此分类中其他 20815 个服务器的 98% 更新
  • 基础检查结果等于或优于此分类中其他 20815 个服务器的 60%

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。

常见问题

StoryOS 安全吗?

它在 5 项适用的基础检查中通过了 4 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。

它需要哪些密钥或代码?

你需要填写:STORYOS_TOKEN

它在我的电脑上运行还是在云端?

两种方式都可以:安装在你的电脑上,或者连接云端版本。

它还在维护吗?

GitHub 仓库最后更新于昨天,最新版本(0.1.2)于昨天进入注册中心。

哪些应用可以使用它?

支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。它通过 URL 连接,因此也能用于接受远程 MCP 的应用,例如网页版 ChatGPT(Business 及以上为完整 MCP;Pro 为只读)。

有什么可以替代它?

同一分类中有:World Monitor, openstatus, Oh My Posh Validator, Strata, Jitsu

功能相近的其他选择

同一分类、其他作者、命令已通过检查。官方的排在前面,再按星标排序。

如何看卡片:↓ 每周下载量i★ 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

World Monitor

Real-time global intelligence on conflict, markets, energy, trade, shipping and cyber. 100 tools.

作者:koala73广泛使用★ 9.1万4天前更新

在云端运行✓ 服务方 · worldmonitor.app

openstatus

Manage monitors, status pages, incidents and maintenances in your openstatus workspace.

作者:openstatusHQ广泛使用★ 92762个月前更新

在云端运行✓ 服务方 · openstatus.dev

Strata

MCP server for progressive tool usage at any scale (see https://klavis.ai)

作者:Klavis-AI正在增长★ 582414个月前更新

在云端运行仓库已归档✓ klavis.ai

Jitsu

Manage Jitsu data pipelines: destinations, streams, connections, functions, live events.

作者:jitsucom广泛使用★ 51364个月前更新

在云端运行✓ 服务方 · jitsu.com
与相似项目对比
本项
StoryOS
World MonitoropenstatusOh My Posh ValidatorStrataJitsu
运行位置云端或安装在电脑上都可以在云端运行,无需安装在云端运行,无需安装在云端运行,无需安装在云端运行,无需安装在云端运行,无需安装
需要密钥是否否否否否
可在 Claude 中按 + 添加否否否否否否
基础检查4/54/44/44/41/44/4
许可证NOASSERTIONAGPL-3.0AGPL-3.0MITApache-2.0MIT
价格登记中未列出费用登记中未列出费用登记中未列出费用登记中未列出费用登记中未列出费用登记中未列出费用

作者的 README

打开 README

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。

@storyos/mcp

The StoryOS Model Context Protocol server — it exposes your StoryOS workspace to AI agents (Claude first; ChatGPT-compatible by design). It's a thin, stateless translator over the same /api/v1 the web app uses, so it can never drift from the product, and every response is scoped and validated server-side.

Design + roadmap: docs/product/mcp-plan.md. Ticket: MN-076.

Tools

Read

ToolWhat it does
get_startedOrientation + a workspace map; with section it returns one reference section (the filter cheat-sheet, icons, working notes, a tool's full notes).
list_workspacesWorkspaces the token can access.
list_databasesDatabases in a workspace.
describe_databaseA database's schema — exact api_names, types, options, relation targets: what a write is keyed by.
searchFull-text record search — turn a name into a real id.
query_recordsFilter / sort / paginate records (structured filter AST).
get_recordOne record in full, by uuid or public number.
get_linksWeb-app URLs for a database, its saved views, and/or a batch of records — without a round-trip per record.

Write (each returns the resulting record; each 422 is surfaced verbatim)

ToolWhat it does
create_recordCreate a record; values by api_name, selects accept the label, relations accept an array of target numbers/ids (linked in the same write).
update_recordMerge-update (null clears); record by uuid or public number. Naming a relation sets it to exactly those targets.
delete_recordTrash a record (restorable 30 days).
link_recordsLink a record to targets through a relation field. replace: true sets the link set to exactly targets — how you re-point or clear a one-to-many link.
unlink_recordsRemove specific links from a relation field (the relation itself stays).
add_commentPost a comment.
attach_fileAttach a file to a record — from a public url (fetched server-side) or inline content_base64; images get a thumbnail.
list_attachments / delete_attachmentList a record's files, or remove one by id.
run_buttonPress a button field, running its automation actions.

Build / schema (agents design the workspace, not just fill it)

ToolWhat it does
list_spaces / create_spaceList / create spaces (a space holds databases).
create_database / update_database / delete_databaseCreate, rename/move, or delete a database (delete needs confirm = name).
add_field / update_field / delete_field / change_field_typeManage fields; select options by label; convert a field's type (dry_run to preview).
create_view / update_view / delete_viewManage views; accepts group_by / card_fields / date fields plus filters + sorts.
create_relation / delete_relationLink two databases (one_to_many / many_to_many) — paired relation fields.
reorder_fields / reorder_viewsSet field / view order by name.

Skills (a workspace's saved skills, over MCP)

ToolWhat it does
list_skillsSkills visible to the caller — their own personal ones, plus every shared one (same visibility rule the in-app Skills list enforces).
run_skillResolve a skill's instructions/when_to_use by name or id, and record the run (last_run_at/last_run_status). The call invokes no model, so you — the connected agent — carry out the instructions against the inputs you pass; they're echoed back, not executed server-side (an agent set to "StoryOS AI" can have the skill named in its Skills field instead).

A workspace's skills are also exposed as native MCP resources and prompts, for clients that understand those primitives (not just tools):

  • Resources: a storyos-skill://{workspace}/{skill} template. resources/list returns one entry per visible skill (across every workspace the token can see); reading one returns the skill rendered as portable Markdown — the exact shape GET /skills/:id/export?format=markdown produces.
  • Prompts: one named prompt per visible skill (e.g. skill_weekly_status_digest), disambiguated by workspace name on a name collision. prompts/get returns the skill's instructions as a single user-role message, ready to paste into context.

Both are read-only and best-effort: a catalog fetch failing for one workspace never blanks out another's, and a failure registering them never breaks the tools (see server.ts).

Automations (build a complete workflow — admin-scoped, #334)

ToolWhat it does
list_automationsThe rules on a database — each with its trigger (field/relation names resolved), condition, actions, enabled state, failure streak and last-run status.
get_automationOne rule's full, editable definition — human-readable database/field/select names alongside their stable ids — ready to feed back into update_automation.
create_automationCreate a rule = trigger (record_created / record_updated / record_linked / schedule / webhook_received) + optional condition (a query_records-style filter) + 1–10 actions (set_values / create_record / add_comment / notify_user / update_linked / send_slack_message / send_webhook / send_email / http_request / run_agent). Field/relation/person names and select labels resolve server-side; the in-app validator (AutomationActionsService.validate) validates the whole rule, so a bad reference comes back as a structured error, not a 500.
update_automationRename, enable/disable, or replace a rule's trigger / condition / actions — every field optional, re-validated the same way.
delete_automationDelete a rule (requires confirm=true); reports the removed rule's name, trigger and action summary.

Automations are an admin surface (the API is @RequiresScope('admin') and creator-gated), so all five tools are advertised only to an admin-scoped token — a read- or write-scoped PAT never sees them and therefore can never mutate a rule. These tools call the same endpoints the in-app RuleEditor does; there is no parallel validation path. Diagnose why a rule did (or didn't) fire with get_runs.

Conveniences: query_records / get_record return select values as labels (not option ids) and rich_text as Markdown; create_record / update_record accept Markdown on a rich_text field (headings, lists, links, code → parsed to blocks) and select labels; create_record reports any unset template fields. get_record / query_records / create_record / update_record all include a url — a clickable web-app link for that record — so an agent never has to describe a manual navigation path.

Hosted Streamable HTTP (for ChatGPT / claude.ai connectors without a local process) is the remaining phase — it lands with the cloud tier (see MN-069).

Light enough for capped clients (ticket #1262)

Every listed tool description is billed on every request of every session, so the listing is kept small: under 60,000 characters in total and no description over 1,500 (src/tool-description-budget.test.ts enforces both against the live registry). A description in the source can be as long as the tool needs; tools/list carries a short lead plus [Full notes: get_started section "TOOL NOTES "] (the one canonical pointer format, double quotes), naming the exact section. get_started with no argument returns only the overview and an index of section names (under 10,000 characters, so no client truncates it); get_started with section set to a name from that index returns that one section: the full text of one tool (the long reference grammars are TOOL NOTES create_automation, install_pack, add_field, create_source), FILTERS, ICONS or WORKING NOTES. Matching is exact first, then case-insensitive; an unknown name returns a text answer listing the valid names. One 85,000-character result was the alternative, and a client that truncates a single result would have silently dropped the grammars. A section is only loaded when called, so it costs a capped client nothing up front (src/get-started-sections.test.ts checks every lead points at a section that exists). Where a lead must say something at call time, it is hand-written with a short: key next to the description. The agent-run profile (?profile=agent-run, the Runner's runs) offers get_started too, because parameter hints (FILTERS, ICONS) name it, but keeps every description WHOLE (WHOLE_DESCRIPTION_PROFILES in src/tool-profiles.ts): its 9 tools total about 3k characters, only ~1k more than leads would cost, and a lead is a round trip. src/profile-pointers.test.ts checks, for every profile, that a pointer names a tool the profile offers and a section its own index lists.

Maintainer-only tools (get_branch_name, get_pinned_tabs, set_pinned_tabs, get_dismissed_hints, set_dismissed_hints, list_icon_set; INTERNAL_TOOLS in src/tool-profiles.ts) are not offered on any customer connection. They appear with ?profile=internal over HTTP, or STORYOS_MCP_PROFILE=internal for a stdio server. The icon names list_icon_set used to return are in get_started section ICONS.

Descriptions state facts and references, never instructions (ticket #1288)

The Claude connector directory policy for tool metadata is that a description contains no instructions about model behaviour or other tools. A tool description says what the tool does, takes and returns; it may state a product fact and carry a neutral reference (See also: describe_database, [Full notes: get_started section "TOOL NOTES "]). Guidance for the model (look at the schema before writing, count with count_records, do not retry a held write, ...) lives in get_started section WORKING NOTES, which is a tool RESULT and not a description. src/tool-description-policy.test.ts enforces this against the live registry: every tool description and every parameter description, on every profile, is scanned for imperatives about another tool, sequencing words around a tool name, "read this before"-style calls to action and behavioural instructions (FORBIDDEN_PATTERNS), with self-tests and an explicit ALLOWED_PHRASES list that needs a reason per entry.

Why it doesn't hallucinate

  • Schema-first: describe_database gives exact api_names; you read before you write.
  • Validation-as-teacher: the API's typed error is surfaced verbatim, so a wrong field/value comes back naming the problem and the model self-corrects in one turn.
  • Structured filters, not a query language: nothing to invent; the op×type matrix ships in get_started section FILTERS.
  • Never invent ids: they come only from search / list_* / a prior result; names and slugs are accepted and resolved server-side.

Configure

1. Get a token — in StoryOS, sidebar → API tokens → create one for the workspace you want. Copy the mn_pat_… (shown once; scoped to that workspace). Without this, the server starts and immediately exits with fatal STORYOS_TOKEN is required — that's the config check working, not a broken install.

2a. Claude Code

claude mcp add storyos \

-e STORYOS_URL=https://app.storyos.dev \ -e STORYOS_TOKEN=mn_pat_xxx \ -- npx -y @storyos/mcp

2b. Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json)

{

"mcpServers": { "storyos": { "command": "npx", "args": ["-y", "@storyos/mcp"], "env": { "STORYOS_URL": "https://app.storyos.dev", "STORYOS_TOKEN": "mn_pat_xxx" } } } }

STORYOS_URL is https://app.storyos.dev for the cloud box, or

只显示了开头部分;其余内容请在仓库中阅读