# aaronjmars/web3-research-mcp [Health: Active]

**Category:** 💰 Finance & Fintech  
**Repository:** https://github.com/aaronjmars/web3-research-mcp  
**GitHub Stars:** 161  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/aaronjmars-web3-research-mcp

## Description
Deep Research for crypto - free & fully local

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

```json
"mcpServers": {
  "web3-research-mcp": {
    "command": "npx",
    "args": ["-y","web3-research-mcp@latest"]
  }
}
```

## Documentation

## What aaronjmars/web3-research-mcp MCP server does

The aaronjmars/web3-research-mcp MCP server provides a structured workflow for researching cryptocurrency tokens and DeFi protocols. It combines general web search with direct lookups against CoinGecko and DeFiLlama, while also supporting source-specific research for services such as CoinMarketCap and news sites.

A research workflow can begin with a plan based on a token name and ticker. The server then supports targeted searches, content retrieval, resource storage, and progress updates. Results can be used to assemble reports that cover market information, technical fundamentals, social or news context, and protocol data.

## How it works

The server creates a research plan with sections for the subject being investigated. Search operations return results for web, news, image, or video queries, and keyword-based searches can be associated with a token. Content fetched from a URL can be saved in text, HTML, Markdown, or JSON form.

Saved results are exposed as resources that can be listed later. Section status can be changed between planned, in-progress, and completed, allowing an agent to track the state of a longer investigation. The repository also defines a `token-research` prompt for initiating comprehensive token research.

Several tools call public data APIs directly rather than relying only on page scraping. CoinGecko tools can resolve ambiguous token matches, retrieve market and supply data, and list active exchange markets. DeFiLlama tools can find protocols and return TVL, fee, chain, token, and fundraising information when available.

## Setup and configuration

The aaronjmars/web3-research-mcp MCP server requires Node.js 16 or newer. Its npm package can be launched through `npx` using the package name `web3-research-mcp` and the `-y` flag.

Claude Desktop and Cursor configurations both use a local stdio entry with the following command and arguments:

```json
{
  "command": "npx",
  "args": ["-y", "web3-research-mcp@latest"]
}
```

The standard CoinGecko and DeFiLlama tools use public API tiers and do not require credentials. Setting `COINGECKO_API_KEY` switches CoinGecko requests to the Pro API and sends the key through the documented request header. The key is optional, not a prerequisite for starting the server.

## Tools and capabilities

Available capabilities include:

- Creating token research plans and updating section status
- Searching the web by query type, including news and media searches
- Searching for and researching a token from a named source
- Fetching URL content and saving it as a resource
- Listing previously saved resources
- Resolving CoinGecko token candidates and retrieving live market data
- Listing CoinGecko trading venues, prices, volumes, spreads, and trade links
- Finding DeFiLlama protocols and retrieving TVL, fees, chains, addresses, and related fields

## Limitations and notes

The aaronjmars/web3-research-mcp MCP server depends on external search services and public APIs, so rate limits can affect search operations. Some websites block automated retrieval, which can cause direct URL fetching to fail with HTTP 403 responses. Search results may also be incomplete and should be checked against the original sources when accuracy matters.

The CoinGecko public tier is described as supporting roughly 30 requests per minute. CoinGecko, CoinGecko ticker, and DeFiLlama requests have a 15-second timeout, while the DeFiLlama protocol index is cached for five minutes per process.

_Full upstream README: https://allmcps.com/mcp/aaronjmars-web3-research-mcp/readme_