Skip to main content
Glama
carlosroman

YAMS (Yet Another MCP Server)

by carlosroman

YAMS

Yet Another MCP Server

YAMS is a batteries-included MCP server that provides the common tools most coding agents need:

  • Web search

  • Fetching and parsing web pages

  • File system access

  • Shell command execution

  • Git integration

  • Basic utility tools

Built in Python. Run with uv.

No Electron. No Docker required. No enterprise platform. Just tools.


Why?

Every agent stack ends up rebuilding the same MCP tools:

  • search

  • fetch

  • grep

  • git

  • run commands

  • read/write files

YAMS packages the boring, standard functionality into a single MCP server that works well locally, in CI, or on remote hosts.

The goal is not to be clever.

The goal is to be useful.


Related MCP server: mcp-devtools

Features

  • Fast startup

  • Single-process architecture

  • Async Python implementation

  • Designed for local-first workflows

  • Works well with Claude Code-style agents

  • Easy to extend with custom tools

  • Sensible defaults

  • Minimal configuration


Installation

From GitHub

uv tool install git+https://github.com/carlosroman/yams.git

Or run without installing:

uvx --from git+https://github.com/carlosroman/yams.git yams

Quick Start

Start the server:

uvx yams

Example MCP client configuration:

{
  "mcpServers": {
    "yams": {
      "command": "uvx",
      "args": ["yams"]
    }
  }
}

Included Tools

Planned tools (none implemented yet):

Tool

Description

Status

search

Search the web

⏳ Planned

fetch

Fetch and parse web pages

⏳ Planned

shell

Execute shell commands

⏳ Planned

read_file

Read files

⏳ Planned

write_file

Write files

⏳ Planned

grep

Search project contents

⏳ Planned

git_status

Git status

⏳ Planned

git_diff

Git diff

⏳ Planned

git_commit

Create commits

⏳ Planned

list_dir

List directories

⏳ Planned

More tools will be added over time.


Philosophy

YAMS follows a few simple rules:

Local-first

Your tools should work on your machine without requiring cloud infrastructure.

Unix-style

Small composable tools beat giant opaque systems.

Predictable over magical

Agents need reliable tooling more than “AI-native” abstractions.

Boring technology wins

Python. JSON. stdio. HTTP.

These things work.


Development

Clone the repository:

git clone https://github.com/your-org/yams.git
cd yams

Install dependencies:

uv sync

Run the server:

uv run yams

Run tests:

uv run pytest

Lint:

uv run ruff check

Format:

uv run ruff format

Project Structure

yams/
├── pyproject.toml
├── src/yams/
│   ├── server.py
│   ├── tools/
│   ├── transport/
│   └── config/
├── tests/
└── README.md

Roadmap

  • Browser automation tools

  • Better sandboxing

  • Remote execution support

  • Tool permissions

  • Persistent sessions

  • Caching

  • Authentication support

  • Plugin system


Non-Goals

YAMS is not:

  • an agent framework

  • an orchestration platform

  • a workflow engine

  • a hosted service

  • an IDE replacement

It is just an MCP server.


Name

Yes, the name is recursive.

No, this probably didn't need to exist.


License

MIT

Related MCP Connectors

Related MCP Servers

  • A
    license
    B
    quality
    C
    maintenance
    A local-first MCP server that provides AI agents with safe codebase access through file discovery, hybrid lexical-semantic search, and project introspection. It features durable local memory and semantic indexing while keeping all data and processing entirely on your local machine.
    74
    47 npm
    6
    MIT
  • A
    license
    B
    quality
    D
    maintenance
    Production-grade MCP server that gives AI agents safe access to your local dev environment: filesystem, databases, processes, and OpenAPI specs.
    15
    28 npm
    3
    MIT