Version 2.0.30

What: A new tool that integrates Language Server Protocol servers to provide code intelligence features directly to Claude.

How to use:

# Claude can now use LSP operations when analyzing code
# The tool works automatically when LSP servers are configured

Supported operations:

• goToDefinition: Find where a symbol is defined
• findReferences: Find all references to a symbol
• hover: Get hover information (documentation, type info) for a symbol
• documentSymbol: Get all symbols (functions, classes, variables) in a document
• workspaceSymbol: Search for symbols across the entire workspace

Configuration: LSP servers can be configured in two ways:

1. Plugin-based configuration: Plugins can define LSP servers via .lsp.json files or in their manifest's lspServers property
2. Project configuration: LSP server settings include command, args, supported languages, transport type (stdio/socket), initialization options, and workspace settings

Requirements:

• All operations require filePath, line (0-indexed), and character (0-indexed) parameters
• LSP servers must be configured for the file type; operations will return an error if no server is available

Details:

• Evidence: LSP tool definition at at1() at line 455338, LSP server manager initialization at line 454707
• Plugin LSP configuration loading at Ig2() at line 454527
• Comprehensive error handling for LSP failures including: lsp-config-invalid, lsp-server-start-failed, lsp-server-crashed, lsp-request-timeout, lsp-request-failed
• LSP URI decoding support at line 205830 (uO0() function)

What: An automated documentation maintenance system that updates project documentation files based on conversations with Claude.

How it works:

• When you work with Claude and learn new information about your codebase, the system can automatically update designated "Magic Doc" files
• Uses a specialized magic-docs agent type that reads existing documentation and incorporates new learnings
• Focuses on high-level architecture, patterns, and entry points rather than exhaustive code details

Philosophy:

• Keeps documentation CURRENT (not a changelog or history)
• Updates information in-place to reflect the current state
• Removes or replaces outdated information
• Emphasizes terseness: "High signal only. No filler words."

magic-docs agent type at line 476091, documentation update prompt generator Aa5() at line 475984

What: A new configuration option allowUnsandboxedCommands that provides policy-level control over the dangerouslyDisableSandbox parameter.

How to use:

// In your settings
{
  "sandbox": {
    "allowUnsandboxedCommands": false  // Enforce sandbox-only mode
  }
}

Details:

• Default: true (maintains backward compatibility)
• When set to false: All commands run in sandbox mode regardless of the dangerouslyDisableSandbox parameter
• Enforcement: The security check at Q9A() at line 249229 validates both the parameter AND the policy setting
• User communication: Claude receives different system prompt instructions based on this setting (lines 208919-208942)
• Evidence: Configuration schema at line 473000, getter function XP8() at line 208521

What changed: Internal tool execution system migrated from function-based approach to a class-based queue system with better concurrent execution support.

Technical details:

• New Qr1 class manages tool execution queue with status tracking ("queued" → "executing" → "completed" → "yielded")
• Intelligent concurrency: Tools marked as concurrency-safe can execute in parallel, while non-concurrent-safe tools block subsequent processing
• Feature-flag controlled: Uses tengu_streaming_tool_execution flag to gradually roll out new architecture
• Backward compatible: Falls back to renamed legacy functions (Ev5() at line 429788) when feature flag is disabled

User impact:

• Better responsiveness when Claude uses multiple tools simultaneously
• More efficient parallel processing of independent operations
• Improved result streaming as tools complete

Qr1 class definition at line 429311, feature flag check at line 429499

What: Enhanced validation and handling of shell command redirections for safer execution.

Details:

• Command parser SP() at line 208476 separates redirections from commands
• Safety validation detects dangerous redirection patterns
• Already existed in v2.0.29 but continues to provide protection in v2.0.30

Function SP() at line 208476 (v2.0.30), existed as whA() at line 204320 (v2.0.29)

The TodoWrite tool documentation remains unchanged between v2.0.29 and v2.0.30. All guidance on when to use the tool, task states, and management practices are identical.

Documentation at lines 204989-205176 (v2.0.30), previously at lines 204886-205072 (v2.0.29)

The ExitPlanMode tool continues to exist with the same functionality as in v2.0.29. This tool prompts users to exit plan mode after planning is complete.

Tool description at line 432519 (v2.0.29), continues in v2.0.30 with updated variable names

• Removed several internal functions related to command history and UI components (non-user-facing)
• Added new internal error handling and validation classes for LSP protocol support
• Tool execution helper functions renamed but functionality preserved for backward compatibility