🔍 Research & Search¶
These recipes combine web search, MCP prompts, memory, and notes into powerful research workflows.
The Research Pipeline¶
A full research session using multiple MCP features together:
I want to do a mini research project on "Agentic AI systems".
Step 1: Search the web for the latest developments in agentic AI
Step 2: Use the analyze_topic prompt to structure the findings
Step 3: Save 3 key insights to memory
Step 4: Add a note summarizing the research in one sentence
Step 5: Finally, use the historical_report prompt for "Artificial Intelligence"
with 3 paragraphs to give me context
Show each step's result clearly.
Quick Research Prompts¶
Technology Comparisons¶
Search the web: "MCP vs OpenAI function calling — key differences"
Then analyze_topic: "AI tool-use protocols"
Current Events + Analysis¶
Deep Dive Workflow¶
I want to deeply understand "Retrieval Augmented Generation (RAG)".
1. Search web for a concise explanation
2. Use analyze_topic: "Retrieval Augmented Generation"
3. Use historical_report: "The history of information retrieval" (2 paragraphs)
4. Save to memory: a one-line summary of what RAG is
Crypto + Market Research¶
Get the current prices for bitcoin, ethereum, solana, and dogecoin.
Then search the web for the latest crypto market news.
Combine both into a brief market snapshot.
Learning & Study Mode¶
I'm learning about the Model Context Protocol.
1. Use explain_weather_concept for "ENSO" to warm up my understanding
of how a concept explanation should be structured
2. Now use analyze_topic: "Model Context Protocol architecture"
3. Search the web for "MCP server examples GitHub"
4. Save a study note: "MCP learning session — covered architecture and examples"
Multi-Source Synthesis¶
I need a comprehensive overview of "Large Language Model safety".
Please:
1. Search the web for recent research on LLM safety
2. Run analyze_topic: "LLM safety and alignment"
3. Run historical_report: "AI safety research" with 3 paragraphs
4. Synthesize all three into a final 200-word summary
5. Save that summary to memory
Chain of thought
Notice how combining tools (web search), prompts (analyze/report), and memory (save/recall) creates a workflow that's more powerful than any single feature alone. This is the real power of MCP.