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

Canli Research

作者:arhancanli版本 0.3.0i★ 6 (整个仓库的星标)i电脑或云端官方网站 ↗
注册名称io.github.arhancanli/canli-research-mcp

安装方法

What it does

Canli Capital's open research record: papers, killed candidates, trial counts, live record.

  • Connects 电脑或云端
  • Account or key No key needed
  • Category 搜索与 AI 记忆

可搭配使用

教 AI 做这类工作的 Skill,可与本项一起使用

timesfm

TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting.

作者:google-research★ 35,6381 个 skill

Skill · 操作手册

Shadowbroker

Open-source intelligence for the global theater. Track everything from the corporate/private jets of the wealthy, and spy satellites, to seismic events in one unified interface. Hook an AI agent up to have it parse through data and find previously unseen correlations. The knowledge is available to…

作者:BigBodyCobain★ 11,6051 个 skill

Skill · 操作手册

google-maps-scraper

scrape data from Google Maps. Extracts data such as the name, address, phone number, website URL, rating, reviews number, latitude and longitude, reviews,email and more for each place

作者:gosom★ 8,2361 个 skill

Skill · 操作手册

为什么排在这里

列表按使用分(满分 100)排序,由以下部分组成

热度19/80GitHub 星标或每周下载量:较大的计 60,较小的计 20
新鲜度12/1230 天内有发布或代码更新得满分
仍在响应8/8上次检查时云端端点有响应
使用分39/100

质量: 基础检查通过 5/5 项(仅展示,不用于排序) · 上升势头: 35 (只用于“上升中”标签)

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 Canli Research。按照 https://mcp.popnx.com/i/arhancanli-canli-research-mcp.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 5/5 项检查 · 你的提问会经过开发者的服务器(canlicapital.com) · ✓ 命令已检查 2026年11月13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Canli Research
    Remote MCP server URL · 地址https://canlicapital.com/mcp/research

    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

    +ConnectorsCanli Research

  4. 试一试

    用 Canli Research 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 Canli Research,或回答中出现 Canli Research 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Canli Research
    Connection · 地址https://canlicapital.com/mcp/research
  3. 在聊天中选择 Canli Research

    开启新聊天,输入前先从插件列表中选择 Canli Research

  4. 试一试

    用 Canli Research 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 Canli Research,或回答中出现 Canli Research 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 canli-research https://canlicapital.com/mcp/research
  3. 在 Claude Code 中输入 /mcp

    应能看到 Canli Research 在列表中并显示已连接

  4. 试一试

    用 Canli Research 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 Canli Research,或回答中出现 Canli Research 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 Canli Research 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 Canli Research,或回答中出现 Canli Research 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 Canli Research 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 Canli Research,或回答中出现 Canli Research 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://canlicapital.com/mcp/research

JSON 设置在下方“开发者专用”中

卡住了?问问正在用 Canli Research 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 canli-research https://canlicapital.com/mcp/research

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "canli-research": {
      "url": "https://canlicapital.com/mcp/research"
    }
  }
}

VS Code (mcp.json)

{
  "servers": {
    "canli-research": {
      "type": "http",
      "url": "https://canlicapital.com/mcp/research"
    }
  }
}

安装在电脑上

Claude Code

claude mcp add --scope user canli-research -- npx -y canli-research-mcp

Claude Desktop (claude_desktop_config.json)

{
  "mcpServers": {
    "canli-research": {
      "command": "npx",
      "args": [
        "-y",
        "canli-research-mcp"
      ]
    }
  }
}

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "canli-research": {
      "command": "npx",
      "args": [
        "-y",
        "canli-research-mcp"
      ]
    }
  }
}

VS Code (mcp.json)

{
  "servers": {
    "canli-research": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "canli-research-mcp"
      ]
    }
  }
}

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
6 (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
690/周
连接方式
电脑或云端
传输方式
stdio, streamable-http
最新版本i
0.3.0
发布到注册中心i
2026年10月13日
代码最后更新i
2026年11月18日
最近检查
2026年11月16日
作者
arhancanli
软件包
npm: canli-research-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

工具名称和描述与服务器在 2026年11月16日 报告的完全一致。这些内容由作者撰写,不是我们写的;我们以纯文本形式显示。

  • search_research

    Find Canli Capital research papers (strategy tests, killed candidates, literature reviews, feasibility protocols) by words in their titles and summaries. Use it to find a paper's slug, then read it with get_paper; to browse by subject instead, use list_topics.

  • list_topics

    The research topics, with how many papers each holds and what it covers.

  • get_paper

    A research paper's text as published, by slug, with its headings; send section for one part, and max_chars to cap the length. Find the slug with search_research or list_topics first.

  • trial_ledger

    How many distinct hypotheses Canli Capital has tried against its declared budget, and how many were killed or survived. Use it to judge any published result against the size of the search behind it.

  • live_record

    The live paper-trading record (returns, costs, risk, corrections, provenance) and each sleeve's paper equity, with their limits. Use it for how the strategies are doing now; for why they exist, read the papers.

  • chain_head

    The head of the tamper-evident chain that shows the published record was not rewritten after publication, with where to verify it. Use it to check that a figure you read was not changed later.

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19日

  • ✓ 通过

    作者与仓库一致

    注册名称已通过 GitHub 账户 arhancanli 验证,该账户拥有此仓库

  • ✓ 通过

    仍在维护

    代码最后更新于昨天

  • ✓ 通过

    有许可证

    使用 MIT

  • ✓ 通过

    端点有响应

    在 299 毫秒内响应(HTTP 200)

  • – 不适用

    使用 OAuth 登录

    无需登录

  • ✓ 通过

    工具描述中没有隐藏指令

    已读取 6 个工具描述,未发现针对 AI 的隐藏指令

已公开的漏洞(OSV.dev)

OSV.dev 中没有 canli-research-mcp@0.3.0 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15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/arhancanli-canli-research-mcp.svg)](https://mcp.popnx.com/en/servers/arhancanli-canli-research-mcp)

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“搜索与 AI 记忆”中的其他服务器比较i

  • 星标数多于此分类中其他 6416 个服务器的 0%
  • 发布到注册中心的时间比此分类中其他 6416 个服务器的 50% 更新
  • 基础检查结果等于或优于此分类中其他 6416 个服务器的 99%

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

Canli Research 安全吗?

它在 5 项适用的基础检查中通过了 5 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?

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(0.3.0)于37天前进入注册中心。

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:OpenSEO, Exa, Microsoft Learn MCP, Docs, COTAL actions (cotal.ai)

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

OpenSEO

Keyword research, site audits, rank tracking and AI visibility for your AI agent.

作者:every-app广泛使用★ 2.8万5天前更新

在云端运行✓ everyapp.com

Exa

Connect AI agents to Exa for web search, content fetching, and multi-step research.

作者:exa-labs广泛使用★ 52442个月前更新

在云端运行✓ 服务方 · exa.aiPopNx 团队已实际试用

Microsoft Learn MCP

Official Microsoft Learn MCP Server – real-time, trusted docs & code samples for AI and LLMs.

作者:MicrosoftDocs广泛使用★ 199712个月前更新

在云端运行✓ 服务方 · microsoft.comPopNx 团队已实际试用

Docs

Search and read Empryo's documentation. Read-only, no auth, no local access.

作者:proxysoul正在增长★ 16262个月前更新

在云端运行✓ empryo.com
与相似项目对比
本项
Canli Research
OpenSEOExaMicrosoft Learn MCPDocsCOTAL actions (cotal.ai)
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
需要密钥否否否否否否
可在 Claude 中按 + 添加否是是是否否
基础检查5/55/52/44/44/44/4
许可证MITMITMITCC-BY-4.0NOASSERTIONApache-2.0
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。

canli-research-mcp

An MCP (Model Context Protocol) server for Canli Capital's open research record. It gives an AI assistant six read-only tools over what canlicapital.com publishes: every research paper and killed candidate, the research topics, the count of hypotheses tried against the declared budget, the live paper-trading record with its sleeves, and the head of the chain that shows the published record was not rewritten.

Every result carries the limits its source states, beside two of its own: these are research documents, each with its own claim boundary, and everything here is paper execution, not funded performance or investment advice.

[canli-research-mcp on canlicapital.com](https://canlicapital.com/mcp-servers/research) · All Canli Capital MCP servers. Free, MIT-licensed. For every server behind one install, use canli-mcp.

Tools

ToolReadsReturns
search_research/research-index.jsonPapers matching words in their titles and summaries, as columns and rows
list_topics/research-index.jsonEach topic, how many papers it holds, and what it covers
get_paper/research/{slug}.mdA paper's text and headings; section returns one part, max_chars caps the length
trial_ledger/api/v1/trials/summary.jsonHypotheses tried against the budget, killed and survived
live_record/api/v1/record.json, /api/v1/sleeves.jsonThe live paper record and each sleeve's paper equity
chain_head/api/v1/chain/head.jsonThe head of the tamper-evident chain, and where to verify it

Prompts and resources

PromptAsks
preflight_ideaHas this quant idea already been tried here, how did it end, and what does the literature say?
audit_canli_recordHypotheses tried against the budget, the live paper record with its limits, and how to verify the chain
literature_reviewEvery paper on a topic, summarized with links
ResourceHolds
canli://schemas/{tool}A tool's exact input and output JSON Schema and description
canli://examples/{language}/{tool}A working call against the hosted endpoint in python, javascript or curl
canli://limitsWhat every result does not establish

Install

claude mcp add canli-research -- npx -y canli-research-mcp

Hosted, no install (claude.ai connectors, ChatGPT, Cursor, or any client that takes a URL):

claude mcp add --transport http canli-research https://canlicapital.com/mcp/research

Any MCP client that spawns a process works the same way: npx -y canli-research-mcp.

Cost and speed

The server reads only static files served from canlicapital.com's CDN, caches each one in memory for the session (ten minutes), and computes nothing, so a repeated question costs no request. The tool list a model re-reads on every turn is small and identical on every launch, which lets model providers serve it from their prompt cache; a test keeps it under 2,500 characters and byte-stable.

Configuration

VariableDefaultMeaning
CANLI_API_BASEhttps://canlicapital.comWhere the record is read from.

Privacy

The server sends only GET requests for public files, with no key, cookie or query of yours; it writes nothing to disk.

Related

Unreleased: bounded response and cache admission

The repository implementation accepts at most 4 MiB of response-body bytes per static file, counted before decoding. It reads the native response stream rather than materializing an unbounded response first. A usable advertised Content-Length above the cap is refused before any read; absent or misleading lengths still require actual byte accounting. HTTP errors are refused before their bodies are read. Valid UTF-8, including code points split across chunks, preserves the published text; invalid or truncated UTF-8 is refused rather than silently replaced.

One existing 15-second request signal covers both fetch and body reads. Known failures attempt reader cancellation and release once; a late fetch response is discarded without reading it. Cancellation promises are observed without extending the request deadline. The transport may already have allocated an oversized supplied chunk, and JavaScript cannot preempt a blocked event loop or synchronous transport code. This is an accepted-body-byte bound and cooperative request deadline, not a total process-memory guarantee or a hosted latency measurement.

Only complete, decoded successful responses enter the ten-minute session cache. JSON is parsed before admission; HTTP errors, overflow, abort, invalid UTF-8 and malformed JSON do not become cache entries. A failed concurrent response cannot evict a newer valid result. Cache ages must be finite, nonnegative and strictly below ten minutes, so clock rollback and expiry require a fresh request. There are no added retries, tools, configuration options or dependencies.

This change is Unreleased. Native Response/ReadableStream fixtures exercise these boundaries with injected responses; their execution and release status are recorded in repository evidence.

The response-bound change has32 new native-stream cases passing in the existing remote MCP job (47 research tests in total), with separate initial source and retained-CI reviews by both independent reviewers. Original source/logs/API captures and the four literal receipts are in the27-member evidence archive. Its exact PRIMARY allocation preserves twenty original assets plus four reviews and three documentary files. Final documentary-head verification and publication remain separate; these repository fixtures do not establish hosted latency, adoption or real financial outcomes.

Unreleased: bounded session cache retention

The repository session cache retains at most 32 entries and 8 MiB of text measured as UTF-8 bytes. A valid cache hit or admission moves that path to the most recent position; admission evicts the least recently used paths until both bounds fit. Replacement releases the old value's weight. Cache operations also remove unrelated entries whose age is negative, nonfinite or at least ten minutes. Clock values and timestamps must be finite numbers; strings and other unusable metadata are not coerced into fresh evidence.

JSON validation still precedes admission. Failed, malformed, HTTP-error or aborted responses cannot remove a newer healthy concurrent result, and a validated success is returned even when unusable admission time prevents retention. The 4 MiB response-byte limit, strict UTF-8 reader, cooperative 15-second request scope and all six tool outputs remain unchanged.

The cache remains a Map for get, has, iteration, set, delete and clear inspection in injected fixtures. Inspection does not promote an entry or expire it. Normal set admission enforces the bounds; controlled lookups remeasure retained text before allowing a hit, so changed test-entry fields cannot leave stale byte accounting. Replacing the session cache, bypassing set with Map.prototype, or using accessor properties and mutation during an operation is outside the supported fixture contract. Direct field mutation can exceed a bound between operations; the next controlled cache operation repairs it before returning a hit.

The bound covers retained text bytes per session, not JavaScript string/object overhead, transport allocations, parsed or returned values, the number of sessions, whole-process memory or hard wall time. There is no autonomous timer, configurable policy, retry or new dependency. Finite native/injected fixtures contain at most 12 MiB of synthetic payload per case; execution, independent review and release status are separate evidence gates. This change is Unreleased and makes no measured hosted latency, token, cost or adoption claim.

The cache-retention change has35 distinct new cases passing once in the existing remote MCP job (82 research cases total), with separate source and retained-CI reviews by both independent reviewers. The fixtures include fresh retained hits and refetch after capacity eviction. Original source/intake/logs/API captures, four literal review receipts and prior actual-delivery custody are in the30-member evidence archive. Its exact PRIMARY allocation includes23 origin assets, four initial reviews and three documentary files. Final documentary-head verification and publication remain separate; these fixtures do not establish hosted latency, adoption, indexing or financial outcomes.