gangtise-mcp
The Gangtise MCP server provides comprehensive access to financial market data, investment research, company fundamentals, and AI-driven insights from the Gangtise OpenAPI platform.
Reference & Search
Query current date/time/timezone (
gangtise_current_date)Look up broker/institution codes, Shenwan industry codes (
gangtise_lookup)Search securities by name, code, or pinyin (
gangtise_securities_search)Search constants (industries, cities, announcement categories), concepts, and sectors with constituents
Research & Opinions
Domestic chief analyst opinions, meeting minutes/summaries, broker research reports
Foreign institutional research reports and opinions, independent overseas analyst opinions
A-share and HK announcements (with downloads)
Events & Schedules
Query roadshows, site visits, strategy meetings, and forums
Market Data
Historical daily K-line for A-shares, HK, US stocks, and indices
Minute-level K-line for A-shares
Real-time market snapshots for A/HK/US stocks
Fundamentals
Income statements, balance sheets, cash flow statements (cumulative/quarterly) for A-shares and HK stocks
Main business breakdown, top 10 shareholders, earnings consensus forecasts, valuation metrics with historical percentiles
AI-Powered Analytics
Semantic knowledge base search across reports, minutes, opinions, and announcements
Investment clues by stock or industry, hot topic briefings (morning/noon/evening)
AI-generated one-pager, investment logic report, peer comparison, research outline, earnings review, bull/bear debate
Management discussion extraction from announcements and earnings calls
Theme/concept tracking reports, concept index info and constituents
Cloud, Voice & Personal Data
Browse/download cloud drive files, voice recording transcripts, conference recordings with ASR/AI summaries
WeChat group messages, personal stock watchlist pools
Alternative Data
Search and retrieve EDB industry economic indicator time-series data
Utilities
Large responses (>256KB) are auto-truncated to temp files; use
gangtise_read_responsefor paginated reading.
Allows querying WeChat group messages and chatroom data through the Gangtise platform.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@gangtise-mcpget the latest quarterly financials for AAPL"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
gangtise-mcp
基于 Gangtise OpenAPI 的 MCP(Model Context Protocol)服务,让 Workbuddy, OpenClaw, Hermes, Cherry Studio, Cursor, Claude, Codex 等 AI 助手直接访问 Gangtise 投研平台数据。
Changelog
README 仅列最近 5 个版本摘要:
0.1.50 — 2026-08-03:同步 CLI v0.30.0–v0.31.0——修复服务端 2026-08-01 重构后三个 EDE 取数工具全线不可用(
universe改名 + 截面矩阵转置 + 日期下沉到每个指标),修正复权参数名adjustType(旧名静默取到不复权数据),新增条件选股工具,并把整行/整列丢数据标成_partial。0.1.49 — 2026-07-26:新增财报日历工具,取数护栏改按实际请求行数判定,
fieldList收成闭集以拦截静默错列。0.1.48 — 2026-07-24:修复无效字段名导致的静默错列(数据污染),并把单票总市值路由到 EDE
qte_mkt_cptl。0.1.47 — 2026-07-24:把 EDE 取数参数配方写进工具描述,并按端点收窄
999999无数据提示。0.1.46 — 2026-07-23:取数路由调整,多证券财务/估值批量优先走 EDE 截面/时序接口。
历史里程碑
0.1.45:同步 CLI v0.28.0,适配新版错误码三层重排、日期严格校验与 traceId 透出。
0.1.44:
server.instructions重写为路由层,建立 92 工具积分目录与自动计费标签,大响应支持字段投影与_available_fields。0.1.36–0.1.43:多轮对抗式审查收口——计费端点
no-replay、429 退避与Retry-After、异步任务截止时间与dataId保全、紧凑 JSON,以及 OIDC 发布链 verify/publish 拆分。0.1.33–0.1.35:确立 loud-partial 契约(分页与分片失败均标记
_partial而非静默空洞),token 缓存改原子写并与 CLI 共享自愈。0.1.31–0.1.32:接入 EDE 证券级数据指标,补齐美股财报与公告、个股看点、首席搜索;全量工具端到端联调。
0.1.28–0.1.30:新增产业公众号资讯,token 服务端失效自愈,CI 加 Node 20/22/24 矩阵与 npm provenance 发布。
0.1.24–0.1.27:日程类工具按 API spec 各自收窄字段,本地静态表迁移到服务端常量/题材/板块接口。
0.1.14–0.1.23:确立大响应截断与
gangtise_read_response续读契约,全工具声明readOnlyHint,日期指引上收到 server instructions。0.1.3–0.1.13:铺开基础工具面——港美股行情与三大报表、EDB 另类数据、自选股池,并落地全市场 K 线分片与超 256KB 落盘预览。
完整更新明细及更早版本见 CHANGELOG.md。
Related MCP server: financial-research-agent
功能覆盖
前置要求
Node.js ≥ 20.18.1(undici 7.27+ 的要求,见
package.json#engines)Gangtise 开放平台账号(申请地址),获取
accessKey/secretKey
快速开始
Claude Code
claude mcp add gangtise \
-e GANGTISE_ACCESS_KEY=your_access_key \
-e GANGTISE_SECRET_KEY=your_secret_key \
-- npx -y gangtise-mcp@latestClaude Desktop
编辑配置文件(根据系统选择路径):
macOS:
~/Library/Application Support/Claude/claude_desktop_config.jsonWindows:
%APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"gangtise": {
"command": "npx",
"args": ["-y", "gangtise-mcp@latest"],
"env": {
"GANGTISE_ACCESS_KEY": "your_access_key",
"GANGTISE_SECRET_KEY": "your_secret_key"
}
}
}
}修改后重启 Claude Desktop 生效。
Cursor
编辑 ~/.cursor/mcp.json(全局)或项目根目录下 .cursor/mcp.json:
{
"mcpServers": {
"gangtise": {
"command": "npx",
"args": ["-y", "gangtise-mcp@latest"],
"env": {
"GANGTISE_ACCESS_KEY": "your_access_key",
"GANGTISE_SECRET_KEY": "your_secret_key"
}
}
}
}Windsurf
编辑 ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"gangtise": {
"command": "npx",
"args": ["-y", "gangtise-mcp@latest"],
"env": {
"GANGTISE_ACCESS_KEY": "your_access_key",
"GANGTISE_SECRET_KEY": "your_secret_key"
}
}
}
}Cline(VS Code 插件)
打开 VS Code → Cline 插件面板 → MCP Servers → Edit MCP Settings,加入:
{
"gangtise": {
"command": "npx",
"args": ["-y", "gangtise-mcp@latest"],
"env": {
"GANGTISE_ACCESS_KEY": "your_access_key",
"GANGTISE_SECRET_KEY": "your_secret_key"
}
}
}其他支持 MCP 的客户端
配置格式通用,只需在对应客户端的 MCP 配置文件中加入:
{
"command": "npx",
"args": ["-y", "gangtise-mcp@latest"],
"env": {
"GANGTISE_ACCESS_KEY": "your_access_key",
"GANGTISE_SECRET_KEY": "your_secret_key"
}
}升级到最新版本
npx -y gangtise-mcp 不会每次都去 registry 拉最新版——npx 会把已下载的版本缓存到 ~/.npm/_npx/<hash>/ 下,后续启动直接复用。npm 发布了新版本但客户端工具列表没出现新工具时,多半就是这个原因。
任选其一:
方法 1:配置里钉版本(推荐) —— 把 args 改成 ["-y", "gangtise-mcp@latest"] 或具体版本 ["-y", "gangtise-mcp@0.x.x"],重启 MCP 客户端即可强制拉新。
方法 2:清 npx 缓存
# macOS / Linux
rm -rf ~/.npm/_npx
# Windows (PowerShell)
Remove-Item -Recurse -Force $env:LOCALAPPDATA\npm-cache\_npx清完缓存后,在 MCP 客户端里关掉再打开 gangtise 服务(或重启客户端),npx 会重新下载最新版。
怎么确认当前跑的是哪个版本?查
~/.npm/_npx/*/node_modules/gangtise-mcp/package.json的version字段。
环境变量
变量 | 默认值 | 说明 |
| — | 开放平台 Access Key(与 SECRET_KEY 配对使用) |
| — | 开放平台 Secret Key |
| — | 直接传 Bearer Token(优先于 Key/Secret,适合临时使用) |
|
| API 基础地址(旧域名 |
|
| 单次请求超时(毫秒) |
|
| 异步 AI 任务默认等待超时(毫秒);保持在 MCP 客户端请求超时(约 60s)以下,超时返回 dataId 供 |
|
| Token 缓存文件路径 |
|
| 分页并发数 |
|
| 工具结果内联字节上限;超过则落盘为临时文件并返回可翻页的预览指针。默认 64KB(约 1.5–2 万 token)控制单次响应体积;批量导出可调大(最低 8192) |
| — | 设为 |
认证优先级:GANGTISE_TOKEN > Token 缓存文件 > GANGTISE_ACCESS_KEY + GANGTISE_SECRET_KEY(自动换取并缓存 Token)。
大响应处理
当单次工具调用返回超过内联阈值(GANGTISE_INLINE_MAX_BYTES,默认 64 KB)时,完整数据会写入系统临时目录下的 gangtise-mcp-* 目录(macOS 实际在 /var/folders/.../T/ 下;JSON 数据为 response.json,文本类为 response.md),MCP 响应改为内联返回前 20 条预览及元数据:
字段 | 说明 |
|
|
| 完整数据的临时文件路径 |
| 完整响应的 UTF-8 字节数 |
| 文件中的总条数 |
| 本次内联返回的条数(最多 20) |
| 续读工具名(固定为 |
| 文件中是否还有未返回的条目 |
| 本地处理建议(server 与客户端共享文件系统时适用) |
| 采样前 20 行得到的顶层字段名,及实际扫描行数;供 |
| 仅当顶层字段超 50 个时出现( |
续读完整数据请调用 gangtise_read_response 工具(传 _saved_to 路径,按 offset/limit 分页;单页同样受 GANGTISE_INLINE_MAX_BYTES(默认 64KB)字节预算约束)——不要依赖客户端直接读文件,Claude Desktop 等无文件读取能力的客户端只能走该工具。若单条内容过大导致 20 条预览本身也超过阈值,则只返回元数据(字段名仍见 _available_fields),_preview_count 为 0(此时 has_more: true 表示数据全部在文件中)。
宽表可用 fields 只取所需列(如 fields: ["tradeDate","close"])——投影在字节预算之前完成,因此每页能装下更多行。部分字段名拼错会以 _unknown_fields 回显并照常返回其余字段,全部拼错才报错并回列可用字段。
gangtise_read_response 每页也受同一字节预算约束:当「信封 + 最小一行」仍超预算(或列表为空但非列表兄弟字段本身超预算)时,仍返回该内容并标 _oversized: true——此时单页已无法再缩小,但 next_offset 照常推进,翻页不会卡死。
_local_hint 仅在 server 与客户端共享文件系统、且客户端获准访问该路径时可用:此时可在本地直接投影/过滤/聚合该文件,只把结果读进上下文。远程 MCP、容器隔离、以及无文件读取能力的客户端(如 Claude Desktop)必须继续走 gangtise_read_response。 注意本地直读不受 MCP 侧 owned-temp-path 校验保护,其安全性取决于客户端自身的文件权限。
开发
git clone https://github.com/gangtiser/gangtise-mcp
cd gangtise-mcp
npm install
npm run dev # 直接运行源码(tsx,无需 build)
npm run build # 编译 TypeScript → dist/
npm test # 运行测试发布维护
本包默认通过 GitHub Actions + npm Trusted Publisher 发布,不在本地执行 npm publish,也不需要长期 npm token。发布前确保 npm 包设置已信任本仓库的 .github/workflows/npm-publish.yml workflow;该 workflow 已配置 permissions: id-token: write,推送 v* tag 后会通过 OIDC 发布到 npm。
标准流程:
npm version patch --no-git-tag-version
# 更新 README Changelog,并完成代码/测试修改
npm test
npx tsc --noEmit
npm run build
git add .
git commit -m "fix: <message>"
git push origin main
git tag v0.1.x
git push origin v0.1.x发布完成后确认:
gh run list --workflow npm-publish.yml --limit 1
npm view gangtise-mcp version如果 GitHub Actions 的 publish 步骤提示 OIDC/trusted publisher 失败,应先检查 npm 包的 Publishing access 设置是否绑定到 gangtiser/gangtise-mcp 和 .github/workflows/npm-publish.yml,不要改回本地 token 发布。
License
MIT
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- AlicenseBqualityCmaintenanceMCP server that provides AI assistants access to stock market data including financial statements, stock prices, and market news through a Model Context Protocol interface.Last updated112,275MIT
- Flicense-qualityCmaintenanceMCP server that exposes stock research tools (fundamentals, news, technicals, analyst ratings) to AI clients, enabling autonomous generation of structured investment briefs.Last updated

siftingio-mcpofficial
AlicenseAqualityAmaintenanceMCP server that provides access to SiftingIO market data, including live prices, SEC filings, OHLCV bars, 13F holdings, market status, and economic calendar tools for AI assistants.Last updated36351MIT- AlicenseBqualityCmaintenanceMCP server that wraps the Theta Data API to provide AI assistants with real-time and historic stock, options, and index data, including OHLC, trades, quotes, Greeks, and more.Last updated59MIT
Related MCP Connectors
MCP server exposing the Backtest360 engine API as tools for AI agents.
MCP server for Gainium — manage trading bots, deals, and balances via AI assistants
Real SEC, 13F, insider, congress & macro data your AI agent can cite. Hosted MCP, 24 tools.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/gangtiser/gangtise-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server