dev-tools

Dev Tools exposes a suite of developer utilities as an MCP server so AI assistants can call commands like hashing, encoding, UUID generation, JSON formatting, and more directly within conversations. The MCP server is intended to be plugged into Claude, Cursor, or any MCP-compatible host via stdio or HTTP transports. You can run it locally via a.DotNet tool-based MCP endpoint or in Docker, then register the server in your agent's MCP configuration. Once registered, you can phrase requests like “hash this text with SHA-256” or “generate 5 UUIDs” and the assistant will route them to the appropriate tool exposed by the MCP server and return structured results.

Prerequisites:

• .NET 10 SDK (for building or using the .NET-based MCP tooling)
• Docker (optional, for containerized MCP server usage)
• Make (optional, for convenience targets)

Install the MCP server (DOTNET tool flow):

1. Install the MCP tool globally: dotnet tool install --global Dev.Tools.Mcp
2. Register the MCP server in your AI agent configuration (example): { "dev-tools": { "type": "stdio", "command": "dev-tools-mcp" } }

Run the MCP server via Docker (HTTP/stdio options):

1. Run the MCP server image with stdio/transport configured as needed: docker run --rm -i -e MCP_TRANSPORT=stdio ghcr.io/dobermanch/dev-tools-mcp
2. Alternatively, run the all-in-one dev-tools container that includes MCP alongside UI/API: docker run -p 80:80 -p 8080:8080 -p 8081:8081 ghcr.io/dobermanch/dev-tools

If you prefer a local build, clone the repo, install dependencies, and build:

1. git clone https://github.com/dobermanch/dev-tools.git
2. cd dev-tools
3. dotnet restore
4. dotnet build
5. dotnet test

Tips and caveats:

• The MCP server integrates with multiple tools automatically; no boilerplate needed because tools are emitted from a single [ToolDefinition] source in the library.
• When using Docker, you can choose between stdio and HTTP transports depending on how your AI host communicates. For stdio, set MCP_TRANSPORT=stdio in the container environment.
• The server exposes a REST/API endpoint when using the separate REST API image; MCP is designed to be used by AI assistants through stdio or HTTP transports.
• Ports: The All-in-one Docker image publishes 80 (Web UI), 8080 (REST API), and 8081 (MCP server HTTP transport). The MCP server itself can also be run standalone via the dedicated ghcr.io/dobermanch/dev-tools-mcp image.
• Prerequisite tool names and commands are case-sensitive, and tool discovery is driven by the single tool definition file within Dev Tools.
• If you encounter transport issues, verify MCP_TRANSPORT is set correctly and that the host can reach the container's exposed port or socket.