home / mcp / desktop tools mcp server

Desktop Tools MCP Server

MCP Desktop Tools

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "andrey-zhuravl-mcp-desktop-tools": {
      "command": "python",
      "args": [
        "-m",
        "mcp_desktop_tools.server"
      ],
      "env": {
        "MCPDT_LOG": "INFO",
        "MCPDT_MAX_WORKERS": "8",
        "MCPDT_SUBPROC_TIMEOUT_MS": "30000",
        "MCPDT_SNAPSHOT_INCLUDE_ENV": "1"
      }
    }
  }
}

You use the MCP server from MCP Desktop Tools to inspect and interact with workspaces through a minimal, JSON-based interface. It provides fast search, project maps, git history summaries, skeleton generation, and quick access to recently used files, all over stdin/stdout or local tooling. This makes it easy to build, analyze, and manage codebases across multiple workspaces with consistent, observable behavior.

How to use

You start the server locally and connect to it with an MCP client or tooling that speaks the MCP protocol. The server exposes a small set of tools that you can invoke from your client: search_text to scan code, git_graph to summarise repository history, repo_map to generate a project map, scaffold to generate project skeletons, open_recent to list recently modified files, and snapshot to compose a holistic JSON artifact containing git history, filesystem stats, and safe environment markers.

How to install

Prerequisites you need before installing and running the server:

Install the MCP server package for local development and usage.

pipx install .  # or python -m pip install -e .[dev]

Additional setup and usage notes

The server runs as a local process. Start it with the following command, then connect a client that speaks MCP JSON lines.

python -m mcp_desktop_tools.server

Configuration and security notes

Workspace configurations live in a workspace definition file and can be overridden by environment variables. The server includes security helpers to ensure requests stay within declared workspace roots and detects certain symlink escape attempts. You can tune performance and observability with per-request caching, a bounded worker pool, and optional stage profiling.

Environment and logging

Control logging verbosity with the MCPDT_LOG variable or the client flag you pass to the CLI you use to talk to the server. You can enable per-tool timing metrics and long-running profiling by using the provided flags.

Tools exposed by the server

The server provides the following tools you can call from your MCP client: search_text, git_graph, repo_map, scaffold, open_recent, and snapshot. Each tool returns structured results that your client can interpret and display or store.

Available tools

search_text

Search code across configured workspaces using ripgrep, returning matches and metrics.

git_graph

Summarise git history for a given relative path, including commit information and optionally file lists.

repo_map

Generate a repository map showing directory structure and key files up to a configured depth.

scaffold

Generate a project skeleton from built-in templates with user overrides.

open_recent

List recently modified files within a workspace, filtered by extensions and limits.

snapshot

Compose a combined JSON artifact that includes git history, filesystem stats, and safe environment markers. Can be logged to MLflow if requested.