v0.4.4

New Tool: find_errors

Quickly finds error lines in log files by matching common error patterns.

Added

• find_errors tool: Quickly finds error lines by matching common error patterns
  • Matches: ERROR, FATAL, CRITICAL, Exception, Traceback, Failed, panic, etc.
  • context_lines parameter: lines to show before/after each error (default: 2)
  • include_warnings parameter: also match WARNING patterns (default: false)
  • max_tokens parameter: limit by token count (default: 4000)
  • Shows line numbers with >>> markers for error lines

Rationale

• AI assistants can quickly diagnose issues without reading entire log files
• Common patterns cover most programming languages and frameworks
• Token-based pagination ensures manageable output size

v0.4.2

New Tools: head_log, tail_log, read_log_range

Added

• head_log tool: Reads the beginning of a log file (like Unix 'head' command)
  • lines parameter: specify exact number of lines to read
  • max_tokens parameter: limit by token count (default: 4000)
• tail_log tool: Reads the end of a log file (like Unix 'tail' command)
  • lines parameter: specify exact number of lines to read
  • max_tokens parameter: limit by token count (default: 4000)
• read_log_range tool: Reads a specific range of lines from a log file
  • start_line and end_line parameters for precise range selection
  • max_tokens parameter: limit by token count (default: 4000)
• All tools show line numbers and file statistics

Rationale

• Quick access to start or end of log files without full pagination setup
• tail_log ideal for checking recent log entries during debugging
• read_log_range enables reading arbitrary sections of log files
• lines parameter provides Unix-like head/tail behavior
• Consistent token-based approach across all tools

v0.4.1

Token-based pagination for get_log_content

Added

• Token-based pagination for get_log_content: Now uses max_tokens (default: 4000, max: 100000) to limit output
• Automatic truncation with guidance to use read_log_paginated for large files
• File size and token count shown in output header

Rationale

• Previously get_log_content returned entire file without limits
• Large files could overwhelm AI context windows
• Consistent token-based approach across all tools

Release v0.4.0

Changed

• Token-based pagination for search_log_file: Now uses max_tokens (default: 4000) instead of max_matches for pagination
• max_matches parameter deprecated but still supported for backward compatibility
• Search results now show mode (token-based vs match-based) and estimated token count
• Improved pagination to respect AI context limits by estimating tokens for each match with its context

Rationale

• Search results can be large when matches include context lines
• Token-based pagination provides more predictable memory usage for AI assistants
• Consistent with read_log_paginated pagination strategy
• Better user experience by avoiding context limit issues

v0.2.2

Release v0.2.2

Changed

• Package renamed to log-inspector-mcp for PyPI compatibility
• Updated GitHub repository URL in package metadata
• Documentation improvements for configuration examples

Installation

pip install log-inspector-mcp

PyPI

https://pypi.org/project/log-inspector-mcp/0.2.2/

Full changelog: https://github.com/thhart/log-mcp/blob/main/CHANGELOG.md

v0.2.1

Release v0.2.1

Changed

• Updated package metadata with author information (Thomas Hartwig)
• Fixed license field format for PyPI compatibility
• Minor documentation improvements

Features (from v0.2.0)

• Multiple log directory support
• 4 inspection tools: list, read paginated, regex search
• Auto-resolves filenames across directories
• Built-in AI guidance for proactive log inspection
• Comprehensive documentation with setup examples for Node.js, Python, Java, Rust, Go

Full changelog: https://github.com/thhart/log-mcp/blob/main/CHANGELOG.md