FPD_get_document_download
Download USPTO petition documents, decisions, and exhibits as PDFs using secure proxy links. Generate browser-accessible URLs for patent documents from the USPTO Final Petition Decisions database.
Instructions
Generate browser-accessible download URL for petition documents (PDFs) via secure proxy.
ALWAYS-ON PROXY (DEFAULT): Proxy server starts with MCP - download links work immediately.
Workflow:
fpd_get_petition_details(petition_id='uuid', include_documents=True) → get documentBag
fpd_get_document_download(petition_id='uuid', document_identifier='ABC123') → get download link
Provide download link to user
CRITICAL RESPONSE FORMAT - Always format with BOTH clickable link and raw URL:
📁 | Raw URL: {proxy_url}
Why both formats?
Clickable links work in Claude Desktop and most clients
Raw URLs enable copy/paste in Msty and other clients where links aren't clickable
Document types:
Petition document: Original petition filed with USPTO
Decision document: Director's final decision
Supporting exhibits: Declarations, prior art, technical documents
Parameters:
petition_id: Petition UUID from search results
document_identifier: Document identifier from documentBag
proxy_port: Optional (defaults to FPD_PROXY_PORT env var or 8081)
generate_persistent_link: Generate 7-day persistent link (default: True)
True: Attempts persistent link via USPTO PFW MCP (works across MCP restarts)
False: Session-based link (works while MCP running, no PFW required)
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| petition_id | Yes | ||
| document_identifier | Yes | ||
| proxy_port | No | ||
| generate_persistent_link | No |
Implementation Reference
- src/fpd_mcp/main.py:992-1326 (handler)The primary handler function implementing the FPD_get_document_download tool. It validates inputs, detects centralized proxy (PFW MCP), generates persistent or session-based download URLs, retrieves document metadata from USPTO API, registers with proxy if applicable, and provides LLM-formatted response with clickable links and guidance.
@mcp.tool(name="FPD_get_document_download") @async_tool_error_handler("document_download") async def fpd_get_document_download( petition_id: str, document_identifier: str, proxy_port: Optional[int] = None, generate_persistent_link: bool = True ) -> Dict[str, Any]: """Generate browser-accessible download URL for petition documents (PDFs) via secure proxy. **ALWAYS-ON PROXY (DEFAULT):** Proxy server starts with MCP - download links work immediately. **Workflow:** 1. fpd_get_petition_details(petition_id='uuid', include_documents=True) → get documentBag 2. fpd_get_document_download(petition_id='uuid', document_identifier='ABC123') → get download link 3. Provide download link to user **CRITICAL RESPONSE FORMAT - Always format with BOTH clickable link and raw URL:** **📁 [Download {DocumentType} ({PageCount} pages)]({proxy_url})** | Raw URL: `{proxy_url}` Why both formats? - Clickable links work in Claude Desktop and most clients - Raw URLs enable copy/paste in Msty and other clients where links aren't clickable **Document types:** - Petition document: Original petition filed with USPTO - Decision document: Director's final decision - Supporting exhibits: Declarations, prior art, technical documents **Parameters:** - petition_id: Petition UUID from search results - document_identifier: Document identifier from documentBag - proxy_port: Optional (defaults to FPD_PROXY_PORT env var or 8081) - generate_persistent_link: Generate 7-day persistent link (default: True) - True: Attempts persistent link via USPTO PFW MCP (works across MCP restarts) - False: Session-based link (works while MCP running, no PFW required)""" try: # Input validation if not petition_id or len(petition_id.strip()) == 0: return format_error_response("Petition ID cannot be empty", 400) if not document_identifier or len(document_identifier.strip()) == 0: return format_error_response("Document identifier cannot be empty", 400) # Handle persistent link generation (requires PFW MCP) if generate_persistent_link: centralized_port = os.getenv('CENTRALIZED_PROXY_PORT', '').lower() if centralized_port and centralized_port != 'none': # PFW centralized proxy detected - forward to PFW for persistent link try: pfw_port = int(centralized_port) logger.info(f"Generating persistent link via centralized USPTO PFW proxy (port {pfw_port})") # Construct persistent link request to PFW proxy # PFW proxy should have an endpoint for generating persistent links # Format: POST http://localhost:8080/persistent-link persistent_link_url = f"http://localhost:{pfw_port}/persistent-link" async with httpx.AsyncClient() as client: response = await client.post( persistent_link_url, json={ "source": "fpd", "petition_id": petition_id, "document_identifier": document_identifier, "expires_days": 7 }, timeout=30.0 ) if response.status_code == 200: result = response.json() return { "success": True, "persistent_download_url": result.get("persistent_url"), "expires_in_days": 7, "note": "Generated via centralized USPTO PFW proxy - works across MCP restarts", "ecosystem_integration": "Using PFW centralized database for persistent links" } else: # PFW proxy doesn't support persistent links yet logger.warning(f"PFW proxy persistent link generation failed: {response.status_code}") # Fall through to immediate link with note except Exception as e: logger.warning(f"Failed to generate persistent link via PFW: {e}") # Fall through to immediate link with note # No centralized proxy or persistent link generation failed # Return helpful message encouraging PFW installation if not centralized_port: return { "success": False, "error": "Persistent links require USPTO PFW MCP for centralized database", "suggestion": "Install USPTO PFW MCP for persistent links, or use immediate links (generate_persistent_link=false)", "immediate_alternative": f"Call this tool with generate_persistent_link=false for session-based download link", "pfw_benefits": [ "7-day persistent encrypted links (work across MCP restarts)", "Centralized proxy server (unified rate limiting)", "Cross-MCP document sharing and caching", "Complete USPTO prosecution + petition workflow" ], "note": "FPD provides immediate downloads only - PFW provides persistent links + enhanced features", "recommendation": "Install both USPTO FPD + PFW MCPs for complete patent lifecycle analysis" } else: # PFW is available but persistent link generation failed - fallback to immediate link logger.info("Persistent link generation not available, falling back to immediate link") # Enhanced proxy port detection with centralized proxy support if proxy_port is None: # Check if centralized proxy is available (and not "none") centralized_port = os.getenv('CENTRALIZED_PROXY_PORT', '').lower() if centralized_port and centralized_port != 'none': proxy_port = int(centralized_port) logger.info(f"Using centralized USPTO proxy on port {proxy_port}") else: # Check FPD_PROXY_PORT first (MCP-specific), then PROXY_PORT (generic) proxy_port = get_local_proxy_port() logger.info(f"Using local FPD proxy on port {proxy_port}") # Start proxy server if not already running (unless using centralized proxy) centralized_port_check = os.getenv('CENTRALIZED_PROXY_PORT', '').lower() if not centralized_port_check or centralized_port_check == 'none': await _ensure_proxy_server_running(proxy_port) else: # Centralized proxy is already running (managed by PFW MCP) logger.info("Using centralized proxy - no local proxy startup needed") # Construct proxy URL (port 8081 to avoid conflict with PFW proxy on 8080) proxy_url = f"http://localhost:{proxy_port}/download/{petition_id}/{document_identifier}" # Also construct direct API URL for reference direct_url = f"{api_client.base_url}/{petition_id}/documents/{document_identifier}" # Get petition details to find document metadata petition_result = await api_client.get_petition_by_id(petition_id, include_documents=True) if "error" in petition_result: return petition_result # Extract from nested structure petition_data = petition_result.get(FPDFields.PETITION_DECISION_DATA_BAG, []) if not petition_data: return format_error_response("Petition data not found", 404) # Get documentBag from first petition in array documents = petition_data[0].get(FPDFields.DOCUMENT_BAG, []) # Find document metadata document_metadata = None for doc in documents: if doc.get(FPDFields.DOCUMENT_IDENTIFIER) == document_identifier: document_metadata = doc break if not document_metadata: return format_error_response( f"Document {document_identifier} not found in petition {petition_id}", 404 ) # Track if centralized proxy registration succeeds centralized_registration_success = False # Register document with centralized proxy if using PFW # Check if CENTRALIZED_PROXY_PORT is set and not "none" centralized_port_env = os.getenv('CENTRALIZED_PROXY_PORT', '').lower() if centralized_port_env and centralized_port_env != 'none': # Extract PDF download URL from document metadata download_options = document_metadata.get(FPDFields.DOWNLOAD_OPTION_BAG, []) pdf_download_url = None for option in download_options: if option.get(FPDFields.MIME_TYPE_IDENTIFIER) == 'PDF': pdf_download_url = option.get(FPDFields.DOWNLOAD_URL) break if pdf_download_url: # Extract metadata for enhanced filename generation petition_mail_date = petition_data[0].get(FPDFields.PETITION_MAIL_DATE) app_number = petition_data[0].get(FPDFields.APPLICATION_NUMBER_TEXT) patent_number = petition_data[0].get(FPDFields.PATENT_NUMBER) doc_description = document_metadata.get(FPDFields.DOCUMENT_CODE_DESCRIPTION_TEXT) doc_code = document_metadata.get(FPDFields.DOCUMENT_CODE) # Generate enhanced filename using local proxy logic enhanced_filename = generate_enhanced_filename( petition_mail_date=petition_mail_date, app_number=app_number, patent_number=patent_number, document_description=doc_description, document_code=doc_code, max_desc_length=40 ) logger.info(f"Generated enhanced filename for PFW registration: {enhanced_filename}") try: # Register FPD document with PFW centralized proxy # Use the already validated centralized_port_env variable pfw_port = int(centralized_port_env) register_url = f"http://localhost:{pfw_port}/register-fpd-document" # Create secure token for document access access_token = mcp_auth.create_document_access_token( petition_id=petition_id, document_identifier=document_identifier, application_number=app_number ) async with httpx.AsyncClient() as client: response_reg = await client.post( register_url, json={ "source": "fpd", "petition_id": petition_id, "document_identifier": document_identifier, "download_url": pdf_download_url, "access_token": access_token, # Secure token instead of raw API key "application_number": app_number, "enhanced_filename": enhanced_filename # Professional filename for downloads }, timeout=5.0 ) if response_reg.status_code == 200: logger.info(f"✅ Successfully registered FPD document with centralized proxy") centralized_registration_success = True else: logger.warning( f"❌ Failed to register document with centralized proxy: HTTP {response_reg.status_code}" ) try: error_detail = response_reg.json() logger.warning(f" Registration error details: {error_detail}") except Exception: logger.warning(f" Response body: {response_reg.text[:500]}") except Exception as e: logger.warning(f"❌ Failed to register document with centralized proxy: {e}") # Implement fallback: if centralized registration failed, use local proxy # Only applies if we actually tried to use centralized proxy (not "none") centralized_port_check = os.getenv('CENTRALIZED_PROXY_PORT', '').lower() if centralized_port_check and centralized_port_check != 'none' and not centralized_registration_success: logger.warning("⚠️ Centralized proxy registration failed - falling back to local FPD proxy") # Start local proxy as fallback local_proxy_port = get_local_proxy_port() await _ensure_proxy_server_running(local_proxy_port) # Update proxy URL to use local proxy proxy_url = f"http://localhost:{local_proxy_port}/download/{petition_id}/{document_identifier}" logger.info(f"🔄 Using local FPD proxy on port {local_proxy_port} for this download") # Determine proxy type for response metadata centralized_port_check = os.getenv('CENTRALIZED_PROXY_PORT', '').lower() proxy_type = "centralized" if (centralized_port_check and centralized_port_check != 'none' and centralized_registration_success) else "local" proxy_port_used = proxy_port if proxy_type == "centralized" else get_local_proxy_port() # Build response with LLM guidance for clickable links response = { "success": True, "petition_id": petition_id, "document_identifier": document_identifier, "proxy_download_url": proxy_url, "direct_url": direct_url, "document_info": document_metadata, # Proxy information for debugging "proxy_info": { "type": proxy_type, "port": proxy_port_used, "status": "centralized_registered" if centralized_registration_success else "local_fallback" }, # NEW: Explicit LLM guidance for proper response formatting "llm_response_guidance": { "critical_requirement": "ALWAYS provide BOTH clickable markdown link AND raw URL", "required_format": f"**📁 [Download {document_metadata.get('documentFileName', 'Document')} ({document_metadata.get('pageCount', 'N/A')} pages)]({proxy_url})** | Raw URL: `{proxy_url}`", "user_expectation": "User requested a download - they need immediate browser access to the PDF", "wrong_response": "Don't just show the raw URL or tool results", "correct_response": "Format as clickable markdown link with document description and page count PLUS raw URL for copy/paste", "explanation": "Clickable link works in Claude Desktop, raw URL enables copy/paste in Msty and other clients where links aren't clickable" }, "access_instructions": { "method": "Proxy server download (recommended) or direct API access", "proxy_url": f"{proxy_url} - Click to download via secure proxy", "proxy_port": proxy_port_used, "proxy_note": f"Proxy handles USPTO API authentication ({proxy_type} proxy on port {proxy_port_used})", "rate_limit": "USPTO allows 5 downloads per 10 seconds", "file_type": "PDF document", "estimated_size": f"{document_metadata.get('pageCount', 'unknown')} pages" }, "llm_guidance": { "next_steps": [ f"Present proxy download URL to user: ** [Download {document_metadata.get('documentFileName', 'Document')} ({document_metadata.get('pageCount', 'unknown')} pages)]({proxy_url})**", "Proxy server is now running on port 8081 (started automatically)", "User can click the link to download PDF directly through secure proxy", "Proxy server handles USPTO API authentication automatically" ], "document_context": { "petition_type": petition_data[0].get(FPDFields.DECISION_PETITION_TYPE_CODE_DESCRIPTION_TEXT, "Unknown"), "decision_outcome": petition_data[0].get(FPDFields.DECISION_TYPE_CODE_DESCRIPTION_TEXT, "Unknown"), "decision_date": petition_data[0].get(FPDFields.DECISION_DATE, "Unknown") } }, # Critical UX reminder "ux_critical": "The user wants this PDF file - make the download link immediately clickable!", # Response validation hints "response_validation": { "check_for_markdown_link": "Response should contain [text](url) format", "check_for_clickable_emoji": "Should start with emoji for visual recognition", "check_for_description": "Link text should describe the document type and page count", "success_pattern": f"** [Download {document_metadata.get('documentFileName', 'DocumentType')} ({document_metadata.get('pageCount', 'N')} pages)](http://localhost:{proxy_port}/download/...)**" } } return response except ValueError as e: logger.warning(f"Validation error in get document download: {str(e)}") return format_error_response(str(e), 400) except httpx.HTTPStatusError as e: logger.error(f"API error in get document download: {e.response.status_code} - {e.response.text}") return format_error_response(f"API error: {e.response.text}", e.response.status_code) except httpx.TimeoutException as e: logger.error(f"API timeout in get document download: {str(e)}") return format_error_response("Request timeout - please try again", 408) except Exception as e: logger.error(f"Unexpected error in get document download: {str(e)}") return format_error_response(f"Internal error: {str(e)}", 500) - src/fpd_mcp/main.py:992-992 (registration)MCP tool registration decorator for FPD_get_document_download.
@mcp.tool(name="FPD_get_document_download") - src/fpd_mcp/main.py:1537-1565 (helper)Helper function to run the proxy server (FastAPI app) which serves the /download/{petition_id}/{document_identifier} endpoint used by the tool's generated URLs.
async def _run_proxy_server(port: int = 8081): """Run the FastAPI proxy server Uses API key from Settings (which may come from secure storage or environment variables) """ try: import uvicorn from .proxy.server import create_proxy_app # Pass API key and port from Settings to proxy server # This allows proxy to work with secure storage (Windows DPAPI) app = create_proxy_app(api_key=settings.uspto_api_key, port=port) config = uvicorn.Config( app, host="127.0.0.1", port=port, log_level="info", access_log=False # Reduce noise in logs ) server = uvicorn.Server(config) logger.info(f"HTTP proxy server starting on http://127.0.0.1:{port}") await server.serve() except Exception as e: global _proxy_server_running _proxy_server_running = False logger.error(f"Proxy server failed: {e}") raise - src/fpd_mcp/main.py:1512-1536 (helper)Helper function that ensures the proxy server is started automatically when the tool is called (always-on or on-demand mode).
async def _ensure_proxy_server_running(port: int = 8081): """Ensure the proxy server is running (auto-start on first download)""" global _proxy_server_running, _proxy_server_task if not _proxy_server_running: logger.info(f"Starting HTTP proxy server on port {port}") # Wrap background task with exception handler async def safe_proxy_runner(): try: await _run_proxy_server(port) except Exception as e: logger.error(f"Proxy server crashed: {e}", exc_info=True) global _proxy_server_running _proxy_server_running = False # Allow graceful degradation - main server continues without proxy _proxy_server_task = asyncio.create_task(safe_proxy_runner()) _proxy_server_running = True # Give the server a moment to start await asyncio.sleep(0.5) logger.info(f"Proxy server started successfully on port {port}")