# Vincentwei1021/agent-toolbox [Health: Active]

**Category:** 🔎 Search & Data Extraction  
**Repository:** https://github.com/Vincentwei1021/agent-toolbox  
**GitHub Stars:** 6  
**npm Downloads (last month):** 14  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/vincentwei1021-agent-toolbox

## Description
Production-ready MCP server providing 13 tools for AI agents: web search, content extraction, screenshots, weather, finance, email validation, translation, news, GeoIP, WHOIS, DNS, PDF extraction, and QR code generation. 1,000 free calls/month, no setup required.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "agent-toolbox": {
    "command": "uvx",
    "args": ["langchain-agent-toolbox"]
  }
}
```

## Documentation

## What Vincentwei1021/agent-toolbox MCP server does

Vincentwei1021/agent-toolbox MCP server gives AI agents a common interface for 13 HTTP-backed utilities. The available operations include DuckDuckGo web search, extracting content from URLs, taking webpage screenshots, checking weather, retrieving stock quotes and exchange rates, validating email addresses, translating between more than 100 languages, looking up IP geolocation, searching news, querying WHOIS data, querying DNS records, extracting text from PDFs, and generating QR codes.

The project is useful for agents that combine research, web inspection, network diagnostics, and lightweight content processing. It can replace separate integrations for each of these functions with one API and a consistent JSON response format, according to the supplied project description.

## How it works

The hosted API exposes the capabilities as POST endpoints under `/v1`. Requests use a bearer token in the `Authorization` header. Examples in the project show search requests sent to `/v1/search`, with the search phrase supplied as JSON. The MCP layer makes the same toolbox available to MCP-compatible clients rather than requiring an agent to construct each HTTP request itself.

The service documents caching periods by endpoint. Search, URL extraction, and news results are cached for five minutes; weather and finance data for 15 minutes; screenshots for one hour; email, translation, GeoIP, WHOIS, and PDF results for one hour; DNS queries for five minutes; and QR results for 24 hours. The project also states that it applies rate limiting.

## Setup and configuration

For local MCP use, the README instructs users to install the published `agent-toolbox-mcp` npm package globally and configure the `agent-toolbox-mcp` executable as the command for an MCP server entry. The documented client examples include Claude Desktop, Cursor, and Windsurf. The repository also includes an MCP quickstart example for these clients.

The hosted API provides a registration request that returns a free API key, with no credit card required for the free plan. The supplied MCP configuration excerpt does not show how that key is passed to the local process, so deployments should consult the project’s current setup documentation before selecting a credential variable or configuration field.

Self-hosting instructions use the repository’s npm build and start scripts. The README also describes a Docker workflow that builds the repository image and publishes port 3100, using a `.env` file for configuration. No specific environment variable names are provided in the supplied material.

## Tools and capabilities

- Search the web and find news articles.
- Extract content from ordinary webpages and PDFs.
- Capture screenshots of webpages.
- Retrieve weather, financial, GeoIP, WHOIS, and DNS information.
- Validate email addresses with MX and SMTP checks.
- Translate text across 100 or more languages.
- Generate QR codes.

## Limitations and notes

Vincentwei1021/agent-toolbox MCP server requires API authentication for the hosted service, and the free plan is limited to 1,000 calls per month with a stated rate limit of 60 calls per minute. Pro and Scale plans increase the monthly allowance and rate limit; all listed plans include the 13 endpoints.

The project’s advertised integrations are endpoint-level capabilities, not guarantees that every upstream data source will return complete or current results. Cached results may also mean that repeated requests do not immediately reflect source changes. The README identifies the project as MIT licensed and provides SDK packages for LangChain and LlamaIndex in addition to the MCP interface.

Vincentwei1021/agent-toolbox MCP server is therefore a fit for broad utility coverage, while a project needing one specialized provider, custom source selection, or provider-specific controls may need a more focused integration.

_Full upstream README: https://allmcps.com/mcp/vincentwei1021-agent-toolbox/readme_