Research Assistant

An AI research assistant with web search, file access, and memory. Great for gathering information, summarizing findings, and creating reports.

Ask ChatGPT Ask Claude.ai Open in Cursor

Difficulty: Intermediate Time to setup: 10 minutes

What It Does

Searches the web for current information
Reads and analyzes documents
Creates research reports
Remembers previous research sessions
Can save findings to files

Features

✅ Web Search: Via Brave Search API
✅ File System: Read/write research files
✅ Memory: Remember research context
✅ Artifacts: Generate PDF reports
✅ Tool Chaining: Automatic multi-step research

Prerequisites

MUXI Server running
OpenAI API key
Brave Search API key (get free at brave.com/search/api)

Setup

# 1. Copy this example
cp -r examples/03-research-assistant my-researcher
cd my-researcher

# 2. Install MCP servers
npm install -g @modelcontextprotocol/server-brave-search
npm install -g @modelcontextprotocol/server-filesystem

# 3. Set API keys
muxi secrets setup
# You'll need:
# - OPENAI_API_KEY
# - BRAVE_SEARCH_API_KEY

# 4. Create research directory
mkdir -p research

# 5. Run locally
muxi dev

Test It

Simple Search

curl -X POST http://localhost:8001/v1/chat \
  -H "Content-Type: application/json" \
  -d '{
    "message": "Research the latest developments in AI agent frameworks in 2025",
    "session_id": "research_session_1"
  }'

Multi-Step Research

curl -X POST http://localhost:8001/v1/chat \
  -H "Content-Type: application/json" \
  -d '{
    "message": "Research AI frameworks, compare top 3, and create a comparison report",
    "session_id": "research_session_2"
  }'

Expected Behavior

The agent will:

Search the web for "AI agent frameworks 2025"
Analyze top results
Identify top 3 frameworks
Search for details on each
Create comparison report
Save as PDF artifact

Response:

{
  "text": "I've researched the top AI agent frameworks...

Key findings:
1. LangChain - Most popular...
2. AutoGPT - Best for autonomy...
3. MUXI - Best for orchestration...",
  "artifacts": [
    {
      "name": "ai_frameworks_comparison.pdf",
      "type": "application/pdf",
      "size": 245678
    }
  ],
  "tools_used": [
    {"name": "brave_web_search", "calls": 4},
    {"name": "file_write", "calls": 1}
  ]
}

Configuration Highlights

MCP Tools

MCP servers are defined in mcp/*.afs files:

# mcp/brave-search.afs
schema: "1.0.0"
id: brave-search
type: command
command: npx
args: ["-y", "@modelcontextprotocol/server-brave-search"]
auth:
  type: env
  BRAVE_API_KEY: "${{ secrets.BRAVE_SEARCH_API_KEY }}"

# mcp/filesystem.afs
schema: "1.0.0"
id: filesystem
type: command
command: npx
args: ["-y", "@modelcontextprotocol/server-filesystem", "./research"]

Tool Chaining

Agent automatically chains tools:

1. brave_web_search("AI frameworks")
   → Gets URLs and summaries
2. brave_web_search("LangChain features")
   → Detailed info on framework 1
3. brave_web_search("AutoGPT capabilities")
   → Detailed info on framework 2
4. file_write("comparison.md")
   → Saves findings

Artifacts

artifacts:
  enabled: true
  python_execution: true  # For creating PDFs
  allowed_packages:
    - reportlab  # PDF generation
    - matplotlib  # Charts

Use Cases

Market Research

"Research the electric vehicle market in Europe, focusing on Tesla, VW, and BYD. Create a market analysis report."

Competitive Analysis

"Research our top 3 competitors, analyze their pricing, features, and customer reviews. Create a comparison table."

Academic Research

"Research recent papers on transformer models published in 2024-2025. Summarize key findings and methodologies."

News Monitoring

"Search for news about [your company/industry] from the past week. Summarize key stories and sentiment."

Customization

Add More Tools

Create additional mcp/*.afs files:

# mcp/puppeteer.afs - Web scraping
schema: "1.0.0"
id: puppeteer
type: command
command: npx
args: ["-y", "@modelcontextprotocol/server-puppeteer"]

# mcp/postgres.afs - Database access
schema: "1.0.0"
id: postgres
type: command
command: npx
args: ["-y", "@modelcontextprotocol/server-postgres"]
auth:
  type: env
  DATABASE_URL: "${{ secrets.DATABASE_URL }}"

Change LLM

Update formation.afs:

llm:
  models:
    - text: "openai/gpt-4-turbo"  # Faster, cheaper
    # or
    - text: "openai/gpt-4o"       # Best reasoning

Deploy to Production

# 1. Install MCP servers on production
ssh production-server
npm install -g @modelcontextprotocol/server-brave-search
npm install -g @modelcontextprotocol/server-filesystem

# 2. Deploy formation
muxi deploy production

# 3. Test
curl https://api.yourcompany.com/research/v1/chat \
  -H "Authorization: Bearer $MUXI_API_KEY" \
  -d '{"message": "..."}'

Next Steps

Add More Tools - Database, APIs, etc.
Code Reviewer - GitHub integration example
Multi-Agent Team - Specialized researchers

Common Issues

"MCP server not found"

# Check MCP server is installed
npx @modelcontextprotocol/server-brave-search --version

# Reinstall if needed
npm install -g @modelcontextprotocol/server-brave-search

"Brave API error: Unauthorized"

Check API key is correct:

muxi secrets get BRAVE_SEARCH_API_KEY

Get new key at: https://brave.com/search/api

"Permission denied: /etc/passwd"

Filesystem MCP trying to access disallowed directory. Update mcp/filesystem.afs:

schema: "1.0.0"
id: filesystem
type: command
command: npx
args: ["-y", "@modelcontextprotocol/server-filesystem", "./research"]
# Only allow research directory

"Artifact generation failed"

Install Python packages:

pip install reportlab matplotlib

We use cookies