Skip to content
Research Memex

AI Environment

MCP Explorer Guide

MCP servers are useful only when they solve a concrete research access problem.

Intermediate · 45–60 min · Cherry Studio 1.5+ or Claude Code

An AI that cannot reach outside its own context is a mind without hands.

Overview

MCP (Model Context Protocol) servers connect an AI client to external tools and data sources. They let the model read files, search the web, and access databases for research tasks.

Benefits:

  • Access your research files directly in AI conversations
  • Search academic databases and repositories
  • Automate repetitive research tasks
  • Create structured thinking processes

Part 1: built-in servers for Cherry Studio

Cherry Studio comes with several built-in servers. We'll install these together in class:

Essential research servers

@cherry/filesystem - access your research files

  • Read papers directly from your computer
  • Upload documents to conversations
  • Navigate folder structures
  • Analyze your research files

@cherry/sequentialthinking - structured problem-solving

  • Break complex questions into steps
  • Track reasoning processes
  • Handles systematic review synthesis and other complex tasks
  • Helps you think more methodically
  • Learn more: Mastering Sequential Thinking with MCP

@cherry/fetch - web content retrieval

  • Pull content from URLs
  • Access online papers and reports
  • Retrieve updated information
  • Access gray literature and other online sources

@cherry/memory - persistent memory

  • Remember information across conversations
  • Build knowledge over time
  • Track project progress
  • Maintain research context for long-term conversations

Part 2: discover more

Exploring the Smithery.ai marketplace

Visit Smithery to discover community-built MCP servers.

Popular categories for researchers:

Academic and research:

  • arxiv - search and access academic preprints
  • pubmed - search medical literature
  • scholar - search Google Scholar
  • citation-tracker - track paper citations

Data and analysis:

  • sqlite - database queries and analysis
  • python - execute data analysis scripts
  • r-stats - statistical computing
  • github - code repository access

Web and information:

  • brave-search - web search with API
  • exa - web search
  • webpage-reader - extract content from URLs
  • news-api - current news and trends

Installation from Smithery

Option 1: guided install:

  1. Visit the Smithery playground at https://smithery.ai/playground.
  2. Click "Add Servers."
  3. Paste the server URL (like https://smithery.ai/server/@username/server-name).
  4. Click "Connect."
  5. Copy the configuration to Cherry Studio.

Option 2: manual configuration:

  1. Find the server on Smithery.ai.
  2. Copy the JSON configuration.
  3. In Cherry Studio, go to Settings → MCP Configuration → Add Custom Server.
  4. Paste the JSON and configure it.

Part 3: success story - my MCP journey

So far this guide has been about installing servers other people built. Here's the other side of it: I built one.

The Lotus Wisdom MCP story

I created Lotus Wisdom MCP with Claude's help. As of July 2026, AI models have reached for it roughly 70,000 times to work through a problem, and Smithery currently ranks it the #3 reasoning-category MCP. See the current listing below for up-to-date numbers.

What it does:

  • Helps with creative problem-solving using wisdom traditions
  • Originally designed for getting unstuck in thinking
  • Effective for coding and complex analysis
  • Researchers, developers, and creatives worldwide use it.

Server details: Lotus Wisdom MCP on Smithery.ai

The lesson: you do not need to be a programmer to create useful tools. I worked with Claude to design, test, and deploy this server. If you can describe a repeated research problem clearly enough, you can turn it into a tool.

What could you create?

Ideas for research-focused MCPs:

  • Literature gap finder
  • Methodology matcher
  • Citation network analyzer
  • Reference formatter
  • Research timeline tracker

Think about:

  • What repetitive tasks slow down your research?
  • What information do you wish was easier to access?
  • What analysis would you do if it were automated?

Part 4: challenge

MCP Explorer mission

Your task: discover and test 2-3 MCP servers relevant to your research.

Step by step:

  1. Browse https://smithery.ai/docs for 15 minutes.
  2. Identify servers that could help with your research tasks:
    • CS students: Try arxiv, github, or scholar servers
    • Business students: Try web search or news servers
    • General: Try fetch or python servers for data analysis
  3. Install at least one server in Cherry Studio.
  4. Test it with your actual research files or questions.
  5. Document what worked and what did not.

Treat failures as evidence. Document what does not work so the class can learn from it.

Troubleshooting common issues

Installation problems

"Server not found"

  • Check that the server URL is correct
  • Ensure you have an internet connection
  • Try refreshing the Smithery.ai page

"Configuration failed"

  • Verify that the JSON format is valid
  • Check for missing API keys (some servers require them)
  • Restart Cherry Studio and try again

"Server not responding"

  • Some servers may be temporarily offline
  • Try a different server first
  • Document the issue for class discussion

Usage problems

"No results returned"

  • Check your query format
  • Try simpler test queries first
  • Verify the server supports your request type

"Error messages"

  • Copy the exact error for troubleshooting
  • Try different input
  • Some servers have usage limits

MCPs used in this research stack

We document 3 MCPs for research:

Sequential Thinking MCP: structured analytical reasoning

  • Step-by-step problem-solving for any AI model
  • Handles complex theoretical analysis
  • Full guide →

Lotus Wisdom MCP: contemplative problem-solving

  • Uses wisdom framework for creative breakthroughs
  • Great when you're stuck or need fresh perspectives
  • We used this to design the Research Memex logo
  • Full guide →

Vox MCP: multi-model AI gateway

  • Access 8+ AI providers from any MCP client
  • Pure passthrough — no system prompt injection
  • Compare models, triangulate findings, stress-test arguments
  • Full guide →

Additional resources

Official documentation:

Community:

  • Share discoveries in class discussions
  • Help troubleshoot peer issues
  • Consider contributing to the MCP community; your discoveries may point to the next useful research tool.
cite this page

Lin, X. (2026). MCP Explorer Guide. Research Memex. https://research-memex.org/docs/implementation/ai-environment-setup/mcp-explorer-guide

@misc{docs-implementation-ai-environment-setup-mcp-explorer-guide-2026,
  author = {Xule Lin},
  title = {MCP Explorer Guide},
  year = {2026},
  howpublished = {\url{https://research-memex.org/docs/implementation/ai-environment-setup/mcp-explorer-guide}},
  note = {ORCID: 0000-0001-7885-4194}
}

one renderingthe source remains