Agent Toolbelt Mcp

io.github.marras0914由 GitHub 账号 marras0914 发布v1.0.0更新于 Oct 24, 2026

19 focused API tools for AI agents: contract review, prompt optimizer, context packing, and more.

已验证STDIO仅桌面网页搜索与抓取金融开发者工具

概览

AI 生成的概览

本地 MCP 服务器，通过 Agent Toolbelt API 提供股票研究与通用实用工具，结合实时行情数据和 AI 综合分析。

可以做什么
  • 研究美股：投资论点、盈利历史、内部人 Form 4 信号、估值、多空论证、个股对比与护城河评级
  • 转换与抽取内容：HTML 与 Markdown 互转、从文本提取邮箱电话网址日期、CSV 转类型化 JSON、抓取网址元数据与网页摘要
  • 生成开发素材：JSON Schema、TypeScript 或 Zod 类型、正则、cron 表达式、模拟 API 响应与依赖 CVE 扫描
  • 处理文本与提示词：token 计数与成本估算、提示词评分改写、上下文打包、文档差异对比与合同条款抽取
  • 提供设计与办公辅助：配色方案、品牌套件、地址规范化、会议行动项与图片元数据清除
适用场景
适合需要助手获取实时美股数据并给出结构化投资分析，或希望把模式生成、文本抽取、token 计数、网页摘要等零散小任务交给同一个托管 API 而不是多个独立工具的场景。
运行要求
通过 npx（Node.js）以 stdio 在本地运行。需要在 AGENT_TOOLBELT_KEY 环境变量中提供 Agent Toolbelt API 密钥，注册时需提交邮箱地址。需要能访问 Agent Toolbelt 服务及其上游金融数据提供方。
安装前请注意
API 密钥属于机密，通过环境变量传入；注册需要邮箱地址。调用按量计费：免费额度每月 250 次，股票分析每次 0.02 美元，实用工具每次 0.001 至 0.005 美元，使用会产生费用。工具输入（包括网址、文档和合同）会发送给第三方服务及其数据提供方。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Agent Toolbelt Mcp，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。

Desktop only，通过 STDIO。 STDIO 服务会启动本地进程，因此需要 SourceWeft 桌面宿主。

其他 MCP 客户端

参照 仓库 中的启动说明。

README

agent-toolbelt-mcp

MCP server for Agent Toolbelt — gives Claude real-time stock research capabilities. Seven AI-powered analysis tools that pull live data from Polygon.io, Finnhub, and Financial Modeling Prep, then synthesize it into structured investment analysis.

Try the valuation snapshot live (no signup): www.agenttoolbelt.live

Step 1 — Get a free API key

The MCP server requires an API key. Free tier is 250 calls/month, no credit card required.

bash
curl -X POST 'https://www.agenttoolbelt.live/api/clients/register?source=npm' \  -H "Content-Type: application/json" \  -d '{"email": "[email protected]"}'

The response includes your atb_... key. Save it — you'll need it in Step 2.

Step 2 — Install

Claude Code

bash
claude mcp add agent-toolbelt \  -e AGENT_TOOLBELT_KEY=atb_your_key_here \  -- npx -y agent-toolbelt-mcp

Claude Desktop

Add to claude_desktop_config.json:

json
{  "mcpServers": {    "agent-toolbelt": {      "command": "npx",      "args": ["-y", "agent-toolbelt-mcp"],      "env": {        "AGENT_TOOLBELT_KEY": "atb_your_key_here"      }    }  }}

Stock Research Tools

Seven tools that give Claude live market data and AI-synthesized analysis. Each call pulls from multiple financial APIs in parallel and returns structured JSON. ~4–5 seconds per call. US-listed equities only (NYSE, NASDAQ, AMEX).

ToolWhat it does
stock_thesisFull investment thesis: bullish/neutral/bearish verdict, 2–3 paragraph analysis, key strengths, risks, valuation read, insider interpretation, what to watch next earnings
earnings_analysisEPS beat/miss history across 5 quarters, revenue trend classification, long-term consistency read, upcoming earnings date
insider_signalForm 4 interpretation: distinguishes meaningful open-market purchases from routine option exercises → strong_buy to strong_sell signal + confidence rating
valuation_snapshotP/E, P/S, EV/EBITDA, FCF yield, ROE → cheap/fair/expensive verdict + specific buy zone
bear_vs_bull3 bull + 3 bear arguments steelmanned with real data, net verdict, key debate question, who the stock suits
compare_stocksHead-to-head 2–3 ticker comparison: winner verdict, per-ticker strengths and concerns, recommendation map by investor goal (growth / value / quality)
moat_analysisBuffett-style competitive moat assessment: wide / narrow / none rating, moat sources (brand, switching costs, network effects, scale), durability outlook, and threats

Example

Once installed, ask Claude:

"Give me a full analysis of NVDA — thesis, earnings quality, insider activity, and whether it's cheap right now."

Claude calls the tools in parallel and synthesizes a complete research note. Here's what stock_thesis returns for NVDA:

Verdict: Bullish One-liner: "Nvidia owns the essential infrastructure for the AI revolution with a defensible software moat, but the valuation demands flawless execution."

Key Strengths:

  • Dominant ~80%+ data center GPU market share
  • CUDA moat creates switching costs and customer lock-in
  • 42 buy / 5 hold / 1 sell analyst consensus

Insider Read: Mixed — two executives bought ~47k shares each in March (positive), offset by routine selling from others.

Watch For Next Earnings: Data center revenue growth rate. Deceleration below 30% YoY would signal the boom is maturing.

Utility Tools

20 additional tools for common agent tasks.

ToolWhat it does
generate_schemaJSON Schema / TypeScript / Zod from a description
extract_from_textExtract emails, phones, URLs, dates, currencies from text
convert_markdownHTML ↔ Markdown conversion
fetch_url_metadataTitle, OG tags, author, favicon from any URL
summarize_web_pageFetch a URL, strip boilerplate, return clean Markdown + AI summary with key points
count_tokensToken counts + cost estimates across 15 LLM models
csv_to_jsonCSV to typed JSON with auto delimiter and type casting
normalize_addressUS address → USPS format with component parsing
generate_color_paletteColor palettes with WCAG scores and CSS variables
build_cronNatural language → cron expression with next-run preview
build_regexNatural language → regex with JS/Python/TS snippets
generate_brand_kitFull brand kit — colors, typography, CSS/Tailwind tokens
optimize_promptScore and rewrite LLM prompts for clarity and specificity
extract_meeting_action_itemsAction items, decisions, and summary from meeting notes
compare_documentsSemantic diff between two document versions
extract_contract_clausesKey clauses + risk flags from contracts
mock_api_responseRealistic mock data from a JSON Schema
audit_dependenciesCVE scan for npm/PyPI packages via OSV database
pack_context_windowPack content into a token budget
strip_image_metadataRemove EXIF/GPS metadata from images

Pricing

  • Free tier: 250 calls/month, no credit card
  • Stock analysis tools: $0.02/call (PAYG)
  • Most utility tools: $0.001–$0.005/call

Going to production

When the agent moves out of dev, a new set of questions shows up. What did it call last night? What arguments did it pass? Who approved the destructive one?

Cordon is an MCP gateway that sits in front of servers like this one. Point your client at Cordon instead of directly at Agent Toolbelt; Cordon forwards every call through and adds:

  • A real-time audit log of every tool invocation (name, arguments, response, latency)
  • Per-API-key policy: which tools each caller can use, under what conditions
  • Slack-based human approvals for tool calls you've flagged as high-risk

From the agent's perspective nothing changes — same tools, same schemas. Free tier covers 250 events/month.

License

MIT

来源：mcp-server/README.md，提交 4c3f84c

工具

0
SourceWeft 尚未在沙箱中启动过该软件包，因此其工具未知。

版本历史

1
  1. v1.0.0最新Sep 16, 2026

评分

已满足适用检查项的 40%

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25 分

    请保持端点能够响应 MCP 握手；SourceWeft 每 30 天探测一次。

  • 已声明认证方式10 分

    请在注册中心条目中声明凭据请求头，或保持端点能够响应探测，以便判定其认证方式。

  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list，或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分

    请在仓库或注册中心条目中声明许可证。