patchwork-deepmind
MCP server for the Behringer DeepMind 12 synthesizer. Gives AI agents real-time control over synth parameters via MIDI NRPN, plus edit-buffer snapshots via SysEx.
README
patchwork-deepmind
MCP server for the Behringer DeepMind 12 synthesizer. Gives AI agents real-time control over synth parameters via MIDI NRPN, plus edit-buffer snapshots via SysEx.
Requirements
- Node.js >= 22
- DeepMind 12 connected via USB-MIDI
- macOS (uses native MIDI)
Install
npm install patchwork-deepmind
MCP client setup
Add to your MCP client config (Claude Desktop, VS Code, etc.):
{
"mcpServers": {
"deepmind12": {
"command": "npx",
"args": ["patchwork-deepmind"]
}
}
}
Tools
| Tool | Description |
|---|---|
set_param |
Set a parameter by name (normalized 0–1, raw integer, or enum label) |
set_params |
Batch-set multiple parameters in one call |
describe_param |
Look up a parameter's NRPN, range, units, enum labels, and notes |
describe_nrpn |
Search/list raw NRPN parameters |
describe_fx_type |
List all 35 FX effect types, or get the full param schema for one (settable names, units, min/max, enums) — without needing a snapshot or loading the effect first |
snapshot_state |
Read current patch state via SysEx edit-buffer dump |
send_nrpn |
Send a raw NRPN message by number and value |
patch_edit_buffer |
Read, patch, and write-back raw edit-buffer bytes (use for FX params when set_param can't reach them) |
Parameter guide skill (recommended)
The MCP tools give an agent the ability to control the synth, but not the knowledge of what sounds good — which parameters interact, what value ranges are musical, or how to approach building a specific type of sound.
The skills/deepmind-parameter-guide/ folder is a portable agent skill that provides this. It's organized as a compact index with drill-down sections by synth area (oscillators, filter, envelopes, LFOs, effects, etc.), so an agent loads only the context it needs.
To install the skill, copy it into your project's .claude/skills/ directory:
# From the cloned repo:
cp -r skills/deepmind-parameter-guide .claude/skills/
# Or from the installed npm package:
cp -r $(npm explore patchwork-deepmind -- pwd)/skills/deepmind-parameter-guide .claude/skills/
The skill is self-contained — no dependencies on this repo.
How it works
The server runs as a stdio-based MCP process — no network involved. Your MCP client (Claude Desktop, VS Code, etc.) spawns it as a subprocess and communicates over stdin/stdout. The server auto-detects the DeepMind's USB-MIDI port on startup and performs a SysEx handshake to confirm the connection.
Example
Once connected, you can talk to your agent naturally:
"Give me a warm pad with slow filter movement and a long reverb tail"
"Make the attack slower and add some chorus"
"Snapshot the current patch so I can see what all the values are"
The agent uses the MCP tools to translate these into NRPN messages and SysEx commands in real time. You hear changes immediately on the synth.
Environment variables
| Variable | Default | Description |
|---|---|---|
MIDI_IN |
auto-detect | MIDI input port index or exact name |
MIDI_OUT |
auto-detect | MIDI output port index or exact name |
MIDI_PORT |
— | Shared hint (partial name) used when MIDI_IN/MIDI_OUT are unset |
MIDI_CH |
0 |
MIDI channel (0–15, where 0 = channel 1) |
Troubleshooting
- Server fails to find MIDI port — Make sure the DeepMind is connected via USB and powered on before starting the server. Verify it appears in macOS Audio MIDI Setup.
- Parameters aren't changing on the synth — Check that the DeepMind is set to receive on the correct MIDI channel (Global Settings → MIDI Channel). The default is channel 1.
- Multiple DeepMinds or other MIDI devices — Use
MIDI_IN/MIDI_OUTenv vars to select the correct port by index or name.
Development
npm install
npm run build
npm test
Contributing
Issues and PRs welcome.
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。