dosbox-mcp
MCP server for driving dosbox-automation from Claude Code and other AI tools.
Install / Use
claude mcp add dosbox-automation -- npx -y github:dosbox-automation/dosbox-mcpIf the server publishes to npm under a different name, use that package instead — check the repo README.
MCP Server
Model Context Protocol server
Quality Score
Category
AutomationSupported Platforms
Our assessment of dosbox-mcp
dosbox-mcp scores 75/100 on our quality scale, 2612th of 2,866 Automation skills we index.
Its MCP Server is 5.2 KB long, well organised into 13 sections with 7 code examples: a solid amount of guidance for an agent.
It has 3 GitHub stars, so there is little community track record yet; judge it on its content.
Maintenance, license and trust
- The repository was last updated 23 days ago, so dosbox-mcp is actively maintained.
- No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
- Its trust signals score 75/100, with 3 cautions from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands.
Automated pattern scan on 2026-11-04. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
dosbox-mcp compared with similar skills
All 4 of these similar skills score higher than dosbox-mcp; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| dosbox-mcp (this skill)by dosbox-automation | 75 | 3 | 23d ago | MCP Server |
| Agent-Reachby Panniantong | 100 | 108.8k | today | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 78.8k | today | CLAUDE.md |
| CowAgentby zhayujie | 100 | 47.8k | today | CLAUDE.md |
| Scraplingby D4Vinci | 100 | 96.5k | today | MCP Server |
Frequently asked questions
- How do I install dosbox-mcp?
- Run
claude mcp add dosbox-automation -- npx -y github:dosbox-automation/dosbox-mcp. The install tabs above show the steps for each supported agent. - Which AI agents does dosbox-mcp work with?
- It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
- Is dosbox-mcp safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. It declares no license and scores 75/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is dosbox-mcp still maintained?
- The repository was last updated 23 days ago, so dosbox-mcp is actively maintained.
Skill content
View source on GitHubdosbox-mcp
MCP server for dosbox-automation: drive a DOS machine from an AI agent.
dosbox-automation is a DOSBox fork with a built-in REST API for automation. This bridge connects any MCP-capable agent (Claude Code, Claude Desktop, or anything that speaks the Model Context Protocol) to a running emulator. The agent can type into DOS programs, read the screen, capture frames and video, run sandboxed Lua scripts, and inspect memory.
Getting started
You need the emulator, this bridge, and an MCP client such as Claude Code or Claude Desktop. The bridge runs on Python 3.11 or newer.
1. Get dosbox-automation
Download from the releases page or build from source. Linux and Windows builds are on the releases page; on macOS you build from source. The manual has the platform-specific setup.
2. Enable the API
Open the primary config file. It lives at
~/.config/dosbox-automation/dosbox-automation.conf on Linux,
%LOCALAPPDATA%\dosbox-automation\ on Windows and
~/Library/Preferences/dosbox-automation/ on macOS. In its
[webserver] section (add the section if the file has none), set:
[webserver]
webserver_enabled = true
webserver_token_file = true
mount_allowed_bases = /home/you/dos
The last line names the folder your DOS programs live in. The emulator
refuses to mount anything outside the folders listed there, so without
it the agent can type MOUNT all day and get nothing. On Windows write
the path as C:\Users\you\dos.
Start dosbox-automation. The DOS prompt appears in its window. To check the API is up, open http://localhost:8386/ in a browser: the API's documentation page answers.
3. Install the bridge
Pick one:
pipx install dosbox-mcp
uvx dosbox-mcp
pip install dosbox-mcp inside a virtual environment works too, but
then the dosbox-mcp command lives in that environment's bin
folder, and step 4 needs its full path.
You never start the bridge by hand. Your MCP client launches it when a conversation begins and stops it afterwards. Run in a terminal, it waits silently for a client that never comes.
4. Register with your MCP client
Claude Code, available in every project directory:
claude mcp add --scope user dosbox -- dosbox-mcp
With uvx instead of an install:
claude mcp add --scope user dosbox -- uvx dosbox-mcp
Claude Desktop: open Settings, then Developer, then Edit Config, and
add the server to claude_desktop_config.json:
{
"mcpServers": {
"dosbox": { "command": "/home/you/.local/bin/dosbox-mcp" }
}
}
Use the full path (which dosbox-mcp, or where dosbox-mcp on
Windows): a desktop app does not see your shell's PATH. For uvx, use
"command": "uvx", "args": ["dosbox-mcp"]. Restart Claude Desktop
afterwards; it reads the file only at start.
The bridge reads the emulator's API token from
~/.config/dosbox-automation/webserver/api_token. On Windows and macOS
the emulator writes the token under its own config folder from step 2,
in a webserver subfolder; point the bridge at that file with the
DOSBOX_TOKEN_FILE environment variable or the token_file setting
(see Configuration).
5. Hello world
Start a new conversation with your agent and ask it something like:
Type DIR in the DOS prompt and tell me what files you see.
The agent calls the bridge tools to type the command, reads the screen
text, and reports back what DOS printed. From here you can mount the
folders you listed in step 2, run and install programs, capture
screenshots, search memory for values, or write Lua scripts that
automate entire workflows. The bridge_status tool confirms the
connection if you want to check first.
Configuration
Optional. Write a commented config file to tune the port, spawn an emulator on demand, or limit what an agent may do:
dosbox-mcp setup --init
The comments in the generated file explain each setting. Environment
variables DOSBOX_API_URL, DOSBOX_API_TOKEN and DOSBOX_TOKEN_FILE
override it. The bridge talks to loopback addresses only; it will not
reach out to a remote machine.
What the agent gets
Tools for session control, screen text and captures, video recording, Lua scripting, keyboard and mouse input, and memory access. You choose how much reach an agent has, from read-only observation to full control.
An open protocol
The contract between this bridge and the emulator is written down in PROTOCOL.md and versioned independently. dosbox-automation is the reference implementation; any DOSBox variant that speaks the protocol works with this bridge.
In the box
Like the games it drives, this one ships with its paper. The feelies/
folder holds the manual, a reference card for every tool, an install
sheet, and a tech-support page.
Requirements
- Python 3.11 or newer
- A running dosbox-automation (0.84-da2 or newer) on the same machine
License
GPL-2.0-or-later. See LICENSE. The protocol specification (PROTOCOL.md) is licensed CC-BY-SA-4.0.
This project is developed with tooled assistance, but tested, reviewed and signed off by a human developer.
Related Skills
Agent-Reach
108.8kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
headroom
78.8kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
CowAgent
47.8kOpen-source personal AI assistant & Agent Harness. Plans tasks, runs tools and skills, self-evolves with memory and knowledge. Multi-agent, multi-model, multi-channel. Lightweight, extensible, one-line install.
Scrapling
96.5k🕷️ The BEST Web Scraping framework that handles everything from a single request to a full-scale crawl! Don't be shy, join here: https://discord.gg/EMgGbDceNQ and follow here for daily tips and tricks: https://x.com/Scrapling_dev
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
