- Fixed health check to use correct /healthz endpoint instead of /health
- Added MCP version (mcpVersion) and supported n8n version (supportedN8nVersion) to health check response
- Added versionNote field with instructions for AI agents about manual version verification
- n8n API limitation: instance version cannot be determined automatically
- Updated axios usage for healthz endpoint access with proper error handling
- Added workflowNodeType field to all node-returning MCP tools
- AI agents now receive both internal format (nodes-base.webhook) and workflow format (n8n-nodes-base.webhook)
- Created getWorkflowNodeType() utility to construct proper n8n format from package name
- Solves issue where AI agents would search nodes and use wrong format in workflows
- No database changes required - uses existing package_name field
- Updated: search_nodes, get_node_info, get_node_essentials, get_node_as_tool_info, validate_node_operation
- Updated CHANGELOG.md with comprehensive documentation of the changes
This completes the fix for issue #71, ensuring AI agents can seamlessly create workflows
with the correct node type format without manual intervention.
- Add centralized normalizeNodeType utility to handle prefix conversion
- n8n-nodes-base.* → nodes-base.*
- @n8n/n8n-nodes-langchain.* → nodes-langchain.*
- Update all 9 affected MCP tools to use normalized node types
- AI agents can now use node types directly from n8n workflow exports
- Maintains backward compatibility with existing shortened prefixes
- Add comprehensive test coverage for all affected methods
Fixes#71🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Remove examples from get_node_essentials responses
- Remove examples from validate_node_operation when errors occur
- Update documentation to reflect removal of examples
- Keep helpful format hints in get_node_for_task (different purpose)
The auto-generated examples were misleading AI agents with incorrect
configurations (e.g., Slack "channel" vs "select" property). Tools
now focus on validation and error messages instead of examples.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Redesigned documentation to be utilitarian and AI-agent focused
- Removed all pleasantries, emojis, and conversational language
- Added concrete numbers throughout (528 nodes, 108 triggers, 264 AI tools)
- Updated all tool descriptions with practical, actionable information
- Enhanced examples with actual return structures and usage patterns
- Made Code node guides prominently featured in overview
- Verified documentation accuracy through extensive testing
- Standardized format across all 30+ tool documentation files
Documentation now optimized for token efficiency while maintaining
clarity and completeness for AI agent consumption.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Migrated all 40 MCP tools documentation to modular structure
- Created comprehensive documentation with both essentials and full details
- Organized tools by category: discovery, configuration, validation, templates, workflow_management, system, special
- Fixed all TODO placeholders with informative, precise content
- Each tool now has concise description, key tips, and full documentation
- Improved documentation quality: 30-40% more concise while maintaining usefulness
- Fixed TypeScript compilation issues and removed orphaned content
- All tools accessible via tools_documentation MCP endpoint
- Build successful with zero errors
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Reduced average description length from 250-450 to 93-129 chars
- Documentation tools now average 129 chars per description
- Management tools average just 93 chars per description
- Moved detailed documentation to tools_documentation() system
- Only 2 tools exceed 200 chars (necessarily verbose)
Also includes search_nodes improvements:
- Fixed primary node ranking (webhook, HTTP Request now appear first)
- Fixed FUZZY mode threshold for better typo tolerance
- Removed unnecessary searchInfo messages
- Fixed HTTP node type case sensitivity issue
This significantly improves AI agent performance by reducing context usage
while preserving all essential information.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Critical fixes based on Claude Desktop feedback:
1. Fixed crypto documentation: require('crypto') IS available despite editor warnings
- Added clear examples of crypto usage
- Updated validation to guide correct require() usage
2. Clarified $helpers vs standalone functions
- $getWorkflowStaticData() is standalone, NOT $helpers.getWorkflowStaticData()
- Added validation to catch incorrect usage (prevents '$helpers is not defined' errors)
- Enhanced examples showing proper $helpers availability checks
3. Fixed JMESPath numeric literal documentation
- n8n requires backticks around numbers in filters: [?age >= `18`]
- Added multiple examples and validation to detect missing backticks
- Prevents 'JMESPath syntax error' that Claude Desktop encountered
4. Fixed webhook data access gotcha
- Webhook payload is at items[0].json.body, NOT items[0].json
- Added dedicated 'Webhook Data Access' section with clear examples
- Created process_webhook_data task template
- Added validation to detect incorrect webhook data access patterns
All fixes based on production workflows TaNqYoZNNeHC4Hne and JZ9urD7PNClDZ1bm
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Root cause: AI agents were placing error handling properties inside `parameters` instead of at node level
Major changes:
- Enhanced workflow validator to check for ALL node-level properties (expanded from 6 to 11)
- Added validation for onError property values and deprecation warnings for continueOnFail
- Updated all examples to use modern error handling (onError instead of continueOnFail)
- Added comprehensive node-level properties documentation in tools_documentation
- Enhanced MCP tool documentation for n8n_create_workflow and n8n_update_partial_workflow
- Added test script demonstrating correct node-level property usage
Node-level properties now validated:
- credentials, disabled, notes, notesInFlow, executeOnce
- onError, retryOnFail, maxTries, waitBetweenTries, alwaysOutputData
- continueOnFail (deprecated)
Validation improvements:
- Detects misplaced properties and provides clear fix examples
- Shows complete node structure when properties are incorrectly placed
- Type validation for all node-level boolean and string properties
- Smart error messages with correct placement guidance
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Fixed issue where Docker images using sql.js adapter returned boolean fields
as strings, causing is_trigger=0 to evaluate as true instead of false.
Changes:
- Added convertIntegerColumns() to sql.js adapter to convert SQLite integers
- Updated server.ts and node-repository.ts to use Number() conversion as backup
- Added test script to verify fix works with sql.js adapter
This fixes webhook, cron, interval, and emailReadImap nodes showing
isTrigger: false in Docker deployments.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Renamed start_here_workflow_guide to tools_documentation for clarity
- Added depth parameter to control documentation detail (essentials/full)
- Converted output from JSON to LLM-friendly plain text format
- Added per-tool documentation capability
- Created two-tier documentation system:
- Essentials: brief info with key parameters and tips
- Full: comprehensive docs with examples and best practices
- Documented 8 commonly used MCP tools
- Removed 380+ lines of unused getWorkflowGuide method
- Fixed duplicate tool definitions
- Updated all documentation references
- Added test script for tools documentation
BREAKING CHANGE: start_here_workflow_guide renamed to tools_documentation
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Implement dynamic n8n API configuration checking
- Remove static config export in favor of lazy getter function
- Fix management tools not being registered when env vars set after startup
- Optimize logger performance with cached environment variables
- Clean up debug logging and remove console.error usage
- Bump version to 2.7.1
This ensures all 38 tools (22 documentation + 16 management) are properly
registered regardless of when environment variables become available.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Commented out console.log statements that were interfering with JSON-RPC communication
- Fixed Claude Desktop errors for list_nodes and list_ai_tools
- All stdout in MCP servers must be valid JSON messages
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Renamed files to remove unnecessary suffixes:
- tools-update.ts → tools.ts
- server-update.ts → server.ts
- http-server-fixed.ts → http-server.ts
- Created version utility to read from package.json as single source of truth
- Updated all imports across 21+ files
- Removed legacy files:
- src/http-server.ts (legacy HTTP server with known issues)
- src/utils/n8n-client.ts (unused legacy API client)
- Added n8n_diagnostic tool to help troubleshoot management tools visibility
- Added script to sync package.runtime.json version
- Fixed version mismatch issue (was hardcoded 2.4.1, now reads 2.7.0 from package.json)
This addresses GitHub issue #5 regarding version mismatch and provides better diagnostics for users.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
BREAKING CHANGES:
- Renamed n8n_update_workflow to n8n_update_full_workflow for clarity
NEW FEATURES:
- Added n8n_update_partial_workflow tool for diff-based workflow editing
- Implemented WorkflowDiffEngine with 13 operation types
- Added transactional updates with two-pass processing
- Maximum 5 operations per request for reliability
- Operations can be in any order - engine handles dependencies
- Added validateOnly mode for testing changes before applying
- 80-90% token savings by only sending changes
IMPROVEMENTS:
- Enhanced tool description with comprehensive parameter documentation
- Added clear examples for simple and complex use cases
- Improved error messages and validation
- Added extensive test coverage for all operations
DOCUMENTATION:
- Added workflow-diff-examples.md with usage patterns
- Added transactional-updates-example.md showing before/after
- Updated README.md with v2.7.0 features
- Updated CLAUDE.md with latest changes
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- NEW: n8n_validate_workflow tool to validate workflows from n8n instance by ID
- Fetches workflow via API and runs comprehensive validation
- Uses existing WorkflowValidator for consistency
- Supports all validation profiles and options
- Updated start_here tool with new workflow lifecycle info
- Updated README with new tool documentation
- Updated Claude Project Setup instructions
This completes the workflow lifecycle management:
discover → build → validate → deploy → execute → monitor
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add node type existence validation that catches invalid types
- Fix critical issue: now catches 'nodes-base.webhook' BEFORE database lookup
- Add smart suggestions for common mistakes (webhook → n8n-nodes-base.webhook)
- Add minimum viable workflow validation (prevents single-node workflows)
- Add empty connection detection for multi-node workflows
- Add helper functions for workflow structure examples and fix suggestions
- Prevent AI agents from creating workflows with question mark nodes
This fixes the exact issue where workflows created with 'nodes-base.webhook'
would show as broken (question marks) in the n8n UI.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added complete workflow lifecycle documentation (Discover → Build → Validate → Deploy → Execute)
- Added new 'n8n_management' topic with comprehensive guide for all 14 management tools
- Documented API requirements, limitations, and best practices
- Added examples showing integration between documentation and management tools
- Included debug script for troubleshooting n8n API authentication
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add comprehensive AI tools guide in start_here_workflow_guide
- New 'ai_tools' topic with detailed AI workflow patterns
- Key insight: ANY node can be used as an AI tool
- $fromAI() expression examples and best practices
- AI tool connection validation guidance
- Update recommended workflow to include validate_node_minimal
- Quick required fields check before full validation
- Performance tips for validation tools
- Enhance Claude Project Setup instructions
- Clear 5-phase workflow process
- Emphasis on AI tool integration
- Validation best practices
- Better tool usage examples
- Add validation_guide section to workflow guide
- When to use each validation tool
- Performance considerations
- AI workflow validation patterns
These updates make it much clearer how to use the MCP tools effectively,
especially for AI workflows where ANY node can be connected as a tool.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Update tool descriptions to clarify ANY node can be used as AI tool
- Add get_node_as_tool_info to available tools in README
- Enhance workflow validation tool descriptions for AI connections
- Update README with v2.5.1 release notes
- Remove redundant _General__Scrape_with_HTTP_tool.json file
- Bump version to 2.5.1 in package.json
The key insight: ANY n8n node can be connected to an AI Agent's tool
port, not just those marked with usableAsTool=true. This update makes
that clear throughout the documentation and tool descriptions.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Phase 2 Professional Validation Features:
1. Validation Profiles:
- minimal: Only required fields
- runtime: Critical errors + security warnings
- ai-friendly: Balanced (default)
- strict: All checks + best practices
2. New Node Validators:
- Webhook: Path validation, response modes, auth warnings
- PostgreSQL: SQL injection detection, query safety
- MySQL: Similar to Postgres with MySQL-specific checks
3. New Tools:
- validate_node_minimal: Lightning-fast required field checking
- Updated validate_node_operation with profile support
4. SQL Safety Features:
- Detects template expressions vulnerable to injection
- Warns about DELETE/UPDATE without WHERE
- Catches dangerous operations (DROP, TRUNCATE)
- Suggests parameterized queries
5. Enhanced Coverage:
- Now supports 7+ major nodes with specific validators
- Flexible validation based on use case
- Professional-grade safety checks
This completes the major validation system overhaul from the original plan.
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
- Add validate_node_operation tool with 80%+ fewer false positives
- Remove deprecated validate_node_config tool
- Add EnhancedConfigValidator with operation context filtering
- Add node-specific validators for Slack, Google Sheets, OpenAI, MongoDB
- Integrate working examples in validation responses
- Add actionable next steps and auto-fix suggestions
- Test shows Slack validation reduced from 45 errors to 1 error\!
BREAKING CHANGE: validate_node_config removed in favor of validate_node_operation
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
- Updated search_templates description to clarify it searches template names/descriptions, NOT node types
- Added clear examples showing the correct way to find templates by node type
- Updated template timeframe references from 6 months to 1 year (399 templates)
- Added TEMPLATE SEARCHING section to quick reference guide
- Made it clear that for node-based search, use list_node_templates instead
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Move TemplateFetcher import to dynamic import in fetchAndUpdateTemplates
- Templates now work everywhere since they just read from database
- axios only loaded when actually fetching new templates
- Fixes Docker runtime error while keeping full template functionality
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
- Use dynamic import for TemplateService to avoid runtime errors
- Skip template initialization when IS_DOCKER=true
- Return friendly messages for template tools in Docker mode
- Templates work from pre-built database, no runtime fetching needed
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
- Add 4 new MCP tools for workflow templates
- Integrate with n8n.io API to fetch community templates
- Filter templates to last 6 months only
- Store templates in SQLite with full workflow JSON
- Manual fetch system (not part of regular rebuild)
- Support search by nodes, keywords, and task categories
- Add fetch:templates and test:templates npm scripts
- Update to v2.4.1
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
- Set environment variables BEFORE imports in stdio-wrapper
- Override ALL console methods to prevent any output
- Update docker-entrypoint.sh to use exec for proper signal handling
- Add fallback if stdio-wrapper.js is missing
- Remove background process handling in stdio mode
This ensures absolutely no log output corrupts the JSON-RPC stream
- Add InitializeRequestSchema handler to MCP server
- Implement stdout flushing for Docker environments
- Create stdio-wrapper for clean JSON-RPC communication
- Update docker-entrypoint.sh to prevent stdout pollution
- Fix logger to check MCP_MODE before level check
These changes ensure the MCP server responds to initialization requests
within Claude Desktop's 60-second timeout when running in Docker.
Fixed the initialization timeout issue with minimal changes:
1. Added stdout flush after server connection to combat Docker buffering
2. Fixed docker-entrypoint.sh to not output to stdout in stdio mode
3. Added process.stdin.resume() to keep server alive
4. Added IS_DOCKER environment variable for future use
5. Updated README to prioritize Docker with correct -i flag configuration
The core issue was Docker's block buffering preventing immediate JSON-RPC
responses. The -i flag maintains stdin connection, and explicit flushing
ensures responses reach Claude Desktop immediately.
Also fixed "Shutting down..." message that was breaking JSON-RPC protocol
by redirecting it to stderr in stdio mode.
Docker is now the recommended installation method as originally intended.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added new 'start_here_workflow_guide' tool for AI agents to get started
- Clarified search_nodes behavior (OR logic for multiple words)
- Simplified list_nodes description with package name clarification
- Updated get_node_info to recommend essentials first
- Made get_property_dependencies description more concise with example
- Removed confusing quick reference comment (now in dedicated tool)
The workflow guide tool provides:
- Recommended workflow patterns
- Search tips and best practices
- Common node references
- Known issues and workarounds
- Performance guidance
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
Based on Claude Desktop evaluation feedback, implemented minimal fixes:
## Day 1 - Deploy & Debug
- Added /version and /test-tools endpoints for deployment verification
- Added debug logging to list_nodes and list_ai_tools
- Fixed version display in health and initialization responses
## Day 2 - Core Fixes
- Fixed multi-word search to handle phrases like "send slack message"
- Added property deduplication to eliminate duplicate webhook/email properties
- Fixed package name mismatch to handle both formats (@n8n/ prefix variations)
## Day 3 - Polish & Test
- Added simple in-memory cache with 1-hour TTL for essentials
- Added documentation fallback when nodes lack documentation
- All features tested and verified working
Total code changes: ~62 lines as planned
No overengineering, just simple focused fixes
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add get_node_essentials tool for 10-20 essential properties only
- Add search_node_properties for targeted property search
- Add get_node_for_task with 14 pre-configured templates
- Add validate_node_config for comprehensive validation
- Add get_property_dependencies for visibility analysis
- Implement PropertyFilter service with curated essentials
- Implement ExampleGenerator with working examples
- Implement TaskTemplates for common workflows
- Implement ConfigValidator with security checks
- Implement PropertyDependencies for dependency analysis
- Enhance property descriptions to 100% coverage
- Add version information to essentials response
- Update documentation with new tools
Response sizes reduced from 100KB+ to <5KB for better AI agent usability.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
The recent commit c4d8c73 renamed the exported constant from
'n8nDocumentationTools' to 'n8nDocumentationToolsFinal' in
tools-update.ts, but the imports in other files were not updated.
This caused TypeScript compilation errors during Docker builds:
- Cannot find name 'n8nDocumentationTools'
- Build failures in GitHub Actions
Fixed by updating imports and usages in:
- src/mcp/server-update.ts (lines 9 and 100)
- src/http-server-fixed.ts (lines 8 and 188)
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Fixed console output interference with stdio JSON-RPC protocol:
- Modified logger to suppress ALL output in stdio mode
- Added DISABLE_CONSOLE_OUTPUT environment variable support
- Updated error handlers to respect stdio mode
- Updated Claude Desktop configuration documentation:
- Added required environment variables for clean stdio communication
- Promoted local installation as recommended method
- Fixed remote connection instructions (removed broken mcp-remote)
- Added troubleshooting section for common issues
- Updated both README.md and docs/README_CLAUDE_SETUP.md
- Environment variables now required for stdio mode:
- MCP_MODE=stdio
- LOG_LEVEL=error
- DISABLE_CONSOLE_OUTPUT=true
- NODE_ENV=production
This ensures clean JSON-RPC communication without console output corruption.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Removed unused legacy files:
- server.ts (legacy server with workflow management)
- tools.ts (legacy tool definitions)
- prompts.ts (unused prompt definitions)
- resources.ts (unused resource definitions)
- server-optimized.ts (unused alternative implementation)
- Updated tools-update.ts with improved MCP tool descriptions:
- Clear, action-oriented descriptions following MCP best practices
- Better parameter documentation with examples
- Emphasized documentation-only nature of the server
- Added technical details (e.g., SQLite FTS5 for search)
The active implementation now only uses server-update.ts and tools-update.ts
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Updated README.md with clear setup instructions and recent updates
- Simplified Claude Desktop setup guide with troubleshooting
- Enhanced HTTP deployment guide for production use
- Streamlined troubleshooting guide with quick fixes
- Added mcp-http-client.js for Node.js 16 compatibility
- Fixed stdio mode console output corruption
Key improvements:
- Clear distinction between local and remote deployment
- Node.js 18+ requirement for mcp-remote clearly documented
- USE_FIXED_HTTP=true prominently featured for v2.3.2
- Production deployment best practices
- Multi-user service considerations
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Root Cause Analysis:
- Express.json() middleware was consuming request stream before StreamableHTTPServerTransport
- StreamableHTTPServerTransport has initialization issues with stateless usage
Two-Phase Solution:
1. Removed all body parsing middleware to preserve raw streams
2. Created http-server-fixed.ts with direct JSON-RPC implementation
Key Changes:
- Remove express.json() from all HTTP server implementations
- Add http-server-fixed.ts that bypasses StreamableHTTPServerTransport
- Implement initialize, tools/list, and tools/call methods directly
- Add USE_FIXED_HTTP=true environment variable to enable fixed server
- Update logging to not access req.body
The fixed implementation:
- Handles JSON-RPC protocol directly without transport complications
- Maintains full MCP compatibility
- Works reliably without stream or initialization errors
- Provides better performance and debugging capabilities
Usage: MCP_MODE=http USE_FIXED_HTTP=true npm start
This provides a stable, production-ready HTTP server for n8n-MCP.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add ConsoleManager to prevent console output interference with StreamableHTTPServerTransport
- Implement SingleSessionHTTPServer with persistent session reuse
- Create N8NMCPEngine for clean service integration
- Add automatic session expiry after 30 minutes of inactivity
- Update logger to be HTTP-aware during active requests
- Maintain backward compatibility with existing deployments
This fixes the "stream is not readable" error by implementing the Hybrid
Single-Session architecture as documented in MCP_ERROR_FIX_PLAN.md
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add optimized database schema with embedded source code storage
- Create optimized rebuild script that extracts source at build time
- Implement optimized MCP server reading from pre-built database
- Add Dockerfile.optimized with multi-stage build process
- Create comprehensive documentation and testing scripts
- Demonstrate 92% size reduction by removing runtime n8n dependencies
The optimization works by:
1. Building complete database at Docker build time
2. Extracting all node source code into the database
3. Creating minimal runtime image without n8n packages
4. Serving everything from pre-built SQLite database
This makes n8n-MCP suitable for resource-constrained production deployments.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
