feat: claude mcp server

[fmunteanu]

Objective

New Anthropic claude MCP server implementation, replacing filesystem MCP server:

• claude:Bash - Execute shell commands with proper quoting, timeouts, and background execution support, following OS permissions.
• claude:BashOutput - Monitor output from background processes with regex filtering.
• claude:Edit - Make precise string replacements in files. Great for targeted changes.
• claude:ExitPlanMode - Transition from planning to implementation phase.
• claude:Glob - Fast file pattern matching (e.g., **/*.js, src/**/*.ts). Essential for finding files by pattern.
• claude:Grep - Powerful search using ripgrep with regex support, context lines, and multiple output modes. Superior to bash grep.
• claude:KillBash - Terminate background shell processes.
• claude:MultiEdit - Apply multiple edits to a single file atomically. Efficient for complex refactoring.
• claude:NotebookEdit - Specialized editing for Jupyter notebooks with cell-level operations.
• claude:Read - Read files with line numbers, supports images, PDFs, and Jupyter notebooks. Perfect for examining code structure and content.
• claude:Task - Launch specialized agents (general-purpose, statusline-setup, output-style-setup) for autonomous multi-step tasks.
• claude:TodoWrite - Create and manage structured task lists for complex multi-step work. Tracks progress systematically.
• claude:Write - Create new files or overwrite existing ones. Requires reading the file first if it exists.

The tools are available in both Claude Code and Code Desktop.

Documentation

Install Claude Code and replace the filesystem MCP server with:

    "claude": {
      "command": "claude",
      "args": [
        "mcp",
        "serve"
      ]
    },

• https://axivo.com/claude/wiki/guide/platform/code/#configuration
• https://axivo.com/claude/wiki/guide/platform/desktop/#configuration

Tool Usage Example

Claude delegated the following task to a sub-agent, in Claude Desktop:

claude:Task(
  subagent_type: "general-purpose", 
  description: "Analyze MCP LSP architecture",
  prompt: "Analyze the MCP LSP project architecture in /Users/floren/github/mcp-lsp/src. I need you to:

1. **Map the component relationships** - How do the 4 main modules (index, mcp, client, config, tool) interact with each other?
2. **Identify key abstractions** - What are the main interfaces and classes, and what design patterns are used?
3. **Find potential improvements** - Look for code quality issues, architectural concerns, or areas for optimization.
4. **Document the LSP→MCP mapping** - How does this bridge convert LSP capabilities to MCP tools?

Use the Read, Glob, and Grep tools systematically to understand the codebase thoroughly. Provide a comprehensive technical analysis with specific code examples and recommendations."
)

Scope

• Bug (resolves an issue)
• Enhancement (improves existing functionality)
• Feature (adds new functionality)
• Documentation (adds or improves documentation)

Impact

• Non-breaking (backwards compatible)
• Breaking (backwards incompatible, impacts end-user)

Checklist

• My code follows the contributing guidelines of this project
• I have performed a self-review of my code
• I have made corresponding changes to the documentation
• My changes generate no new issues or warnings