Build an AI Research Assistant with Claude Code and MCP Tools

An AI research assistant isn’t a single tool — it’s a combination of capabilities working together. Web search for finding sources, transcript extraction for video content, news aggregation for current events, and a language model to synthesize everything. Claude Code with the right MCP tools gives you exactly this.

Why This Matters

Research takes time because it involves multiple steps across multiple tools. You search Google, open tabs, read articles, watch videos, take notes, and then try to synthesize everything. Each context switch costs time and attention.

An AI research assistant built on Claude Code collapses these steps. You describe what you’re researching, and the agent handles the searching, reading, and synthesis. It calls tools to gather information and returns a coherent summary — all within your terminal session.

Setup

Install the AgentPatch CLI (zero dependencies, Python 3.10+):

pip install agentpatch

Set your API key:

export AGENTPATCH_API_KEY=your_api_key

Then use it:

ap search "web search"
ap run agentpatch google-search --input '{"query": "test"}'

Get your API key from the AgentPatch dashboard.

Run this command to add AgentPatch as an MCP server:

claude mcp add -s user --transport http agentpatch https://agentpatch.ai/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

Replace YOUR_API_KEY with your actual key from the AgentPatch dashboard. Claude Code discovers all AgentPatch tools automatically.

Building a Research Workflow

Step 1: Web Search for Background

Start with a broad research question:

“Research the current state of WebAssembly adoption in production environments. Find recent articles and data points.”

Claude Code calls the web search tool, finds relevant sources, and summarizes the key findings. You get a structured overview without opening a browser.

Step 2: Video Content Extraction

Conference talks and video tutorials often contain information that isn’t published as text anywhere:

“Find YouTube talks about WebAssembly performance benchmarks from the last 6 months. Get the transcripts and extract the benchmark numbers.”

Claude Code searches for relevant videos, fetches their transcripts through AgentPatch, and pulls out the specific data you asked for.

Step 3: News Monitoring

For topics that are evolving quickly, current news matters:

“Search Google News for WebAssembly announcements in the last month. List any new runtime releases or major adoption stories.”

Claude Code calls the news tool and returns a summary of recent developments.

Step 4: Synthesis

The real value of an AI research assistant is combining all of this:

“Based on everything we’ve gathered about WebAssembly, write a 500-word briefing covering current adoption, performance characteristics, and key players. Include specific data points from the sources.”

Claude Code uses all the information from the previous steps to produce a coherent research document.

Chaining It All Together

You don’t have to run these steps one at a time. You can give Claude Code the full research task:

“I need a research brief on how AI agents are being used in software development. Search the web for recent articles, find 2-3 relevant conference talks on YouTube and get their transcripts, check Google News for recent announcements, and then write a 1-page summary with key findings.”

Claude Code calls multiple tools across the session and produces the final output. The agent handles the research pipeline — you just describe what you need.

Wrapping Up

Building an AI research assistant doesn’t require a custom application. Claude Code with AgentPatch tools handles the common research workflow: search, gather, synthesize. Visit agentpatch.ai to connect the tools and start using Claude Code for research.