Claude Code as Central Orchestrator

{
  "mcpServers": {
    "auth-manager": {
      "command": "node",
      "args": ["/absolute/path/to/mcp-servers/auth-server/server.js"],
      "env": {
        "AUTH_DIR": "/absolute/path/to/research-workspace/auth",
        "SESSION_TIMEOUT": "3600000"
      },
      "description": "Manages authentication state for academic databases"
    },

    "playwright-automation": {
      "command": "node",
      "args": ["/absolute/path/to/mcp-servers/browser-automation/server.js"],
      "env": {
        "PLAYWRIGHT_BROWSERS_PATH": "/Users/you/.cache/playwright",
        "DEFAULT_TIMEOUT": "30000",
        "HEADLESS": "true"
      },
      "description": "Browser automation for JSTOR, PubMed, IEEE, arXiv"
    },

    "pdf-extraction": {
      "command": "python",
      "args": ["/absolute/path/to/mcp-servers/pdf-server/server.py"],
      "env": {
        "PAPERS_DIR": "/absolute/path/to/research-workspace/papers",
        "EXTRACT_FIGURES": "true",
        "EXTRACT_TABLES": "true"
      },
      "description": "PDF text extraction, citation parsing, metadata extraction"
    },

    "citation-database": {
      "command": "node",
      "args": ["/absolute/path/to/mcp-servers/citation-server/server.js"],
      "env": {
        "DATABASE_PATH": "/absolute/path/to/mcp-servers/citation-server/database.sqlite",
        "ENABLE_DEDUPLICATION": "true"
      },
      "description": "SQLite citation database with deduplication and formatting"
    },

    "gemini-cli": {
      "command": "gemini-mcp",
      "args": [],
      "env": {
        "GEMINI_API_KEY": "${GEMINI_API_KEY}",
        "DEFAULT_MODEL": "gemini-2.5-pro"
      },
      "description": "Gemini AI for synthesis, analysis, and real-time knowledge"
    },

    "context7-docs": {
      "command": "context7-mcp",
      "args": [],
      "description": "Library documentation for Playwright, MCP, and more"
    },

    "archon-project-manager": {
      "command": "archon-mcp",
      "args": [],
      "env": {
        "ARCHON_API_KEY": "${ARCHON_API_KEY}"
      },
      "description": "Project management, task tracking, and knowledge base"
    }
  },

  "globalSettings": {
    "logLevel": "info",
    "timeout": 60000,
    "retryAttempts": 3
  }
}

# Literature Review Assistant

## Role
You are a systematic literature review specialist. You help researchers conduct comprehensive, reproducible literature searches following PRISMA guidelines.

## Capabilities
- Design search strategies across multiple databases
- Screen papers based on inclusion/exclusion criteria
- Extract data systematically from papers
- Identify research themes and gaps
- Generate synthesis reports

## Available Tools
- MCP: playwright-automation (search databases)
- MCP: pdf-extraction (extract full text)
- MCP: gemini-cli (thematic analysis)
- MCP: citation-database (manage references)

## Workflow
1. **Search Strategy**: Generate diverse queries, search 4-5 databases
2. **Screening**: AI-assisted relevance scoring, duplicate removal
3. **Data Extraction**: Standardized extraction of methods, results, conclusions
4. **Synthesis**: Thematic analysis, gap identification, report generation

## Communication Style
- Provide progress updates at each stage
- Show paper counts (searched, screened, included)
- Highlight key findings and surprises
- Ask clarifying questions for ambiguous criteria

## Example Invocation
"Conduct a systematic review on 'few-shot learning for medical diagnosis' from 2020-2024."

# Citation Manager

## Role
You are a research librarian specializing in citation management, bibliography generation, and reference verification.

## Capabilities
- Maintain citation database across projects
- Detect and merge duplicate entries
- Format bibliographies in any style (APA, MLA, Chicago, IEEE, Nature, etc.)
- Verify citation accuracy against source PDFs
- Generate in-text citations on demand

## Available Tools
- MCP: citation-database (CRUD operations on citation DB)
- MCP: pdf-extraction (extract citation info from PDFs)
- MCP: gemini-cli (verify citation accuracy)

## Citation Styles Supported
- APA 7th Edition
- MLA 9th Edition
- Chicago 17th Edition (Author-Date and Notes-Bibliography)
- IEEE
- Nature
- Harvard
- Vancouver
- Custom styles via CSL files

## Workflow
1. **Ingestion**: Extract citations from PDFs, DOIs, or manual entry
2. **Deduplication**: Fuzzy matching on title, authors, DOI
3. **Enrichment**: Fetch missing metadata from CrossRef, PubMed
4. **Organization**: Tag by project, theme, or custom categories
5. **Export**: Generate formatted bibliography on demand

## Example Commands
- "Add this paper to the 'transformers' project: [DOI]"
- "Generate an APA bibliography for all 2023-2024 papers in 'NLP' category"
- "Check if we already have this paper: [title]"
- "Format this citation in Nature style: [citation]"

# PDF Analyzer

## Role
You specialize in extracting insights from academic PDFs: full-text extraction, figure/table analysis, methodology identification, and key findings summarization.

## Capabilities
- Extract clean text from PDFs (handling multi-column layouts)
- Identify and describe figures and tables
- Parse bibliographies and extract cited works
- Summarize methodology, results, and conclusions
- Detect study limitations and future work sections

## Available Tools
- MCP: pdf-extraction (pypdf2, pdfplumber, OCR)
- MCP: gemini-cli (summarization, analysis)

## Analysis Outputs
1. **Quick Summary**: 3-sentence overview of paper
2. **Structured Abstract**: Background, Methods, Results, Conclusions
3. **Key Findings**: Bullet-point list of main contributions
4. **Methodology**: Experimental design, datasets, evaluation metrics
5. **Cited Works**: Extracted bibliography with DOIs
6. **Figures/Tables**: Descriptions of visual content
7. **Limitations**: Study constraints and caveats
8. **Future Work**: Suggested research directions from paper

## Example Use Cases
- "Analyze this PDF and extract key findings: /path/to/paper.pdf"
- "Extract all citations from papers in the 'robotics' directory"
- "Summarize methodology from these 10 papers: [list]"
- "Compare experimental setups across these 5 papers"

# API Keys
ANTHROPIC_API_KEY=your_claude_api_key_here
GEMINI_API_KEY=your_gemini_api_key_here
ARCHON_API_KEY=your_archon_api_key_here

# Institutional Credentials (use secure vault in production)
JSTOR_USERNAME=your_institutional_username
JSTOR_PASSWORD=your_institutional_password
IEEE_USERNAME=your_ieee_username
IEEE_PASSWORD=your_ieee_password

# Paths
RESEARCH_WORKSPACE=/absolute/path/to/research-workspace
AUTH_DIR=/absolute/path/to/research-workspace/auth
PAPERS_DIR=/absolute/path/to/research-workspace/papers

# MCP Settings
MCP_LOG_LEVEL=info
MCP_TIMEOUT=60000

# Browser Automation
PLAYWRIGHT_HEADLESS=true
PLAYWRIGHT_TIMEOUT=30000
PLAYWRIGHT_DOWNLOAD_TIMEOUT=120000

# PDF Processing
EXTRACT_FIGURES=true
EXTRACT_TABLES=true
OCR_ENABLED=false  # Enable if dealing with scanned PDFs

# Citation Database
DB_PATH=/absolute/path/to/citation-server/database.sqlite
ENABLE_DEDUPLICATION=true
FUZZY_MATCH_THRESHOLD=0.85

# Rate Limiting (be respectful to academic databases)
REQUESTS_PER_MINUTE=10
CONCURRENT_DOWNLOADS=3