文档管理AI服务器

# Clone and setup
git clone https://github.com/Hellblazer/dt-mcp.git
cd dt-mcp
npm install

# Verify installation
npm run test:scripts

search_devonthink          # Document search with DEVONthink syntax
read_document              # Document content and metadata
create_document            # New document creation
list_databases             # Database enumeration
update_tags                # Tag management
get_related_documents      # AI-suggested relations
create_smart_group         # Dynamic collections
ocr_document               # Text extraction from PDFs/images

advanced_search            # Full search syntax with operators (AND, OR, NOT, field:value)
list_smart_groups          # Access to DEVONthink's organizational features

build_knowledge_graph      # Visual relationship mapping with depth control
find_shortest_path         # BFS pathfinding between documents
detect_knowledge_clusters  # AI-powered document clustering
find_connections           # Multi-type relationship discovery
compare_documents          # Document similarity analysis

automate_research          # Complete research workflows
organize_findings          # Performance-optimized result organization
create_collection          # Research project collections

analyze_document           # Complexity & readability analysis
analyze_document_similarity # Multi-document comparison (performance-optimized)
batch_read_documents       # Parallel document processing

synthesize_documents       # Multi-document synthesis (performance-optimized)
extract_themes             # AI theme identification
classify_document          # Native DEVONthink AI classification
get_similar_documents      # AI-powered similarity finding
create_multi_level_summary # Tiered summarization (brief/detailed/full)
track_topic_evolution      # Topic change analysis over time
create_knowledge_timeline  # Chronological knowledge mapping
identify_trends            # Trending topic detection

batch_search               # Multiple searches in parallel
batch_read_documents       # Multiple document reads

create_collection          # Document collections/research threads
add_to_collection          # Add documents to collections

bulk_import_urls           # Import multiple URLs concurrently with progress tracking
bulk_download_papers       # Download academic papers in bulk with metadata extraction
create_research_project    # Create comprehensive research project structures
execute_workflow           # Execute predefined research workflows
monitor_operations         # Monitor active operations and system resources
manage_operation_queue     # Manage operation queue with priority control

get_tool_help              # AI-friendly help system with examples

1. "Search for 'quantum computing' papers from 2023"
   → advanced_search with date:2023 AND quantum AND computing

2. "Build a knowledge graph from the top result" 
   → build_knowledge_graph with depth 3

3. "Create a research collection"
   → create_collection for "Quantum Research Project"

4. "Synthesize the top 5 documents"
   → synthesize_documents with consensus approach

5. "What topics are trending?"
   → identify_trends across databases

# Boolean operators
"quantum AND physics OR mathematics"

# Field searches  
"name:quantum tag:physics comment:important"

# Wildcards and fuzzy
"quant* ~quantum"

# Exact phrases
"\"quantum computing\""

# Date ranges
"date:2023 AND kind:pdf"

# Find document clusters
detect_knowledge_clusters → Groups by AI themes

# Shortest path between concepts
find_shortest_path from_uuid to_uuid → Connection chain

# Document similarity
get_similar_documents uuid → AI-ranked related documents

# Smart groups
list_smart_groups → All organizational smart groups

# Bulk operations
bulk_import_urls ["url1", "url2", "url3"] → Import multiple URLs concurrently
bulk_download_papers [{source: "arxiv", id: "2301.00001"}, {...}] → Download papers

# Research project setup  
create_research_project "Quantum AI Research" → Full project structure + initial sources

# Workflow orchestration
execute_workflow "academic_research" {topic: "quantum computing"} → Multi-step automation

# System monitoring
monitor_operations → Real-time progress, resource usage, active operations
manage_operation_queue "pause" → Control concurrent operations

{
  "error": "Descriptive error message",
  "details": {
    "tool": "tool_name", 
    "timestamp": "ISO-8601",
    "context": "additional_context"
  }
}

{
  "status": "success",
  "data": { /* tool-specific data */ },
  "metadata": {
    "timestamp": "ISO-8601",
    "tool": "tool_name"
  }
}

Claude ↔ MCP Server ↔ DEVONthink 4 AI
   ↑         ↑              ↑
Request   Thin Wrapper   Native AI

# Quick validation (30 seconds)
npm run test:scripts

# Comprehensive testing (2-5 minutes)  
npm run test:comprehensive

# Individual tool testing
npm run test:tool synthesize_documents '{"documentUUIDs": ["UUID1", "UUID2"], "synthesisType": "summary"}'

# Interactive exploration
npm run test:mcp