Serial Mcp Server — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Serial Mcp Server (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
<!-- mcp-name: io.github.es617/serial-mcp-server -->
A stateful serial port Model Context Protocol (MCP) server for developer tooling and AI agents. Works out of the box with Claude Code, VS Code with Copilot, and any MCP-compatible runtime. Communicates over stdio and uses pyserial for cross-platform serial on macOS, Windows, and Linux.
Example: Let Claude Code list available serial ports, connect to your microcontroller, reset it via DTR, and read the boot banner from your hardware.
Video walkthrough — connecting to a serial device, sending commands, reading responses, and creating plugins.
If you’ve ever copy-pasted commands into screen or minicom, guessed baud rates, toggled DTR to kick a bootloader, and re-run the same test sequence 20 times — this is for you.
You have a serial device. You want an AI agent to talk to it — open a port, send commands, read responses, debug protocols. This server makes that possible.
It gives any MCP-compatible agent a full set of serial tools: listing ports, opening connections, reading, writing, line-oriented I/O, control line manipulation — plus protocol specs and device plugins, so the agent can reason about higher-level device behavior instead of just raw bytes.
The agent calls these tools, gets structured JSON back, and reasons about what to do next — without you manually typing commands into a terminal for every step.
What agents can do with it:
pip install serial-mcp-server
# Register the MCP server with Claude Code
claude mcp add serial -- serial_mcpThen in Claude Code, try:
"List available serial ports and connect to the one on /dev/ttyUSB0 at 115200 baud."
<p align="center"><img src="https://raw.githubusercontent.com/es617/serial-mcp-server/main/docs/assets/scan.gif" alt="Scanning serial ports" width="600"></p>
Once connected, the agent has full serial capabilities:
The agent can coordinate multi-step flows automatically — e.g., toggle reset, wait for prompt, send init sequence, stream output.
At a high level:
Raw Serial → Protocol Spec → Plugin
You can start with raw serial tools, then move up the stack as your device protocol becomes understood and repeatable.
# Editable install from repo root
pip install -e .
# Or with uv
uv pip install -e .MCP is a protocol — this server works with any MCP-compatible client. Below are setup instructions for the most common ones.
# Standard setup
claude mcp add serial -- serial_mcp
# Or run as a module
claude mcp add serial -- python -m serial_mcp_server
# Enable all plugins
claude mcp add serial -e SERIAL_MCP_PLUGINS=all -- serial_mcp
# Enable specific plugins only
claude mcp add serial -e SERIAL_MCP_PLUGINS=mydevice,ota -- serial_mcp
# Debug logging
claude mcp add serial -e SERIAL_MCP_LOG_LEVEL=DEBUG -- serial_mcpAdd to your project's .vscode/mcp.json (or create it):
{
"servers": {
"serial": {
"type": "stdio",
"command": "serial_mcp",
"args": [],
"env": {
"SERIAL_MCP_PLUGINS": "all"
}
}
}
}Adjust env to match your needs — set SERIAL_MCP_PLUGINS to specific plugin names, or add SERIAL_MCP_MIRROR for PTY mirroring.
Add to your project's .cursor/mcp.json (or create it). Cursor does not support dots in tool names, so SERIAL_MCP_TOOL_SEPARATOR must be set to _:
{
"mcpServers": {
"serial": {
"command": "serial_mcp",
"args": [],
"env": {
"SERIAL_MCP_PLUGINS": "all",
"SERIAL_MCP_TOOL_SEPARATOR": "_"
}
}
}
}| Variable | Default | Description |
|---|---|---|
SERIAL_MCP_MAX_CONNECTIONS | 10 | Maximum simultaneous open serial connections. |
SERIAL_MCP_PLUGINS | disabled | Plugin policy: all to allow all, or name1,name2 to allow specific plugins. Unset = disabled. |
SERIAL_MCP_MIRROR | off | PTY mirror mode: off, ro (read-only), or rw (read-write). macOS and Linux only. |
SERIAL_MCP_MIRROR_LINK | /tmp/serial-mcp | Base path for PTY symlinks. Connections get numbered: /tmp/serial-mcp0, /tmp/serial-mcp1, etc. |
SERIAL_MCP_LOG_LEVEL | WARNING | Python log level (DEBUG, INFO, WARNING, ERROR). Logs go to stderr. |
SERIAL_MCP_TRACE | enabled | JSONL tracing of every tool call. Set to 0, false, or no to disable. |
SERIAL_MCP_TRACE_PAYLOADS | disabled | Include write data in traced args (stripped by default). |
SERIAL_MCP_TRACE_MAX_BYTES | 16384 | Max payload chars before truncation (only applies when TRACE_PAYLOADS is on). |
SERIAL_MCP_TOOL_SEPARATOR | . | Character used to separate tool name segments. Set to _ for MCP clients that reject dots in tool names (e.g. Cursor). |
| Category | Tools |
|---|---|
| Serial Core | serial.list_ports, serial.open, serial.close, serial.connection_status, serial.read, serial.write, serial.readline, serial.read_until, serial.flush, serial.set_dtr, serial.set_rts, serial.pulse_dtr, serial.pulse_rts |
| Introspection | serial.connections.list |
| Protocol Specs | serial.spec.template, serial.spec.register, serial.spec.list, serial.spec.attach, serial.spec.get, serial.spec.read, serial.spec.search |
| Tracing | serial.trace.status, serial.trace.tail |
| Plugins | serial.plugin.template, serial.plugin.list, serial.plugin.reload, serial.plugin.load |
Specs are markdown files that describe a serial device's protocol — connection settings, message format, commands, and multi-step flows. They live in .serial_mcp/specs/ and teach the agent what the byte stream means.
Without a spec, the agent can still open a port and exchange data. With a spec, it knows what commands to send, what responses to expect, and what the data means.
You can create specs by telling the agent about your device — paste a datasheet, describe the protocol, or just let it explore and document what it finds. The agent generates the spec file, registers it, and references it in future sessions. You can also write specs by hand.
Plugins add device-specific shortcut tools to the server. Instead of the agent composing raw read/write sequences, a plugin provides high-level operations like mydevice.read_temp or ota.upload_firmware.
The agent can also generate Python plugins (with your approval). It explores a device, writes a plugin based on what it learns, and future sessions get shortcut tools — no manual coding required.
To enable plugins:
# Enable all plugins
claude mcp add serial -e SERIAL_MCP_PLUGINS=all -- serial_mcp
# Enable specific plugins only
claude mcp add serial -e SERIAL_MCP_PLUGINS=mydevice,ota -- serial_mcpEditing an already-loaded plugin only requires serial.plugin.reload — no restart needed.
Every tool call is traced to .serial_mcp/traces/trace.jsonl and an in-memory ring buffer (last 2000 events). Tracing is on by default — set SERIAL_MCP_TRACE=0 to disable.
Two events per tool call:
{"ts":"2025-01-01T00:00:00.000Z","event":"tool_call_start","tool":"serial.read","args":{"connection_id":"s1"},"connection_id":"s1"}
{"ts":"2025-01-01T00:00:00.050Z","event":"tool_call_end","tool":"serial.read","ok":true,"error_code":null,"duration_ms":50,"connection_id":"s1"}connection_id is extracted from args when presentdata is stripped from traced args by default (enable with SERIAL_MCP_TRACE_PAYLOADS=1)Use serial.trace.status to check config and event count, and serial.trace.tail to retrieve recent events — no need to read the file directly.
When the MCP server owns a serial port, most OSes prevent any other process from opening it. PTY mirroring creates a virtual clone port that external tools (screen, minicom, logic analyzers, custom scripts) can connect to simultaneously.
# Enable read-only mirror
claude mcp add serial \
-e SERIAL_MCP_MIRROR=ro \
-- serial_mcp
# After opening a connection, the response includes the mirror path:
# { "mirror": { "pty_path": "/dev/ttys004", "link": "/tmp/serial-mcp0", "mode": "ro" } }
# In another terminal:
screen /tmp/serial-mcp0 115200| Mode | Behavior |
|---|---|
off | No mirror (default). Only the MCP server can access the port. |
ro | External tools see all serial data but cannot write to the device. |
rw | External tools can both see data and write to the device. |
Platform: macOS and Linux only. On Windows, setting SERIAL_MCP_MIRROR to ro/rw logs a warning and is silently ignored.
You can test the server interactively using the MCP Inspector — no Claude or other agent needed:
npx @modelcontextprotocol/inspector python -m serial_mcp_serverOpen the URL with the auth token from the terminal output. The Inspector gives you a web UI to call any tool and see responses in real time.
This server connects an AI agent to real hardware. That's the point — and it means the stakes are higher than pure-software tools.
Plugins execute arbitrary code. When plugins are enabled, the agent can create and run Python code on your machine with full server privileges. Review agent-generated plugins before loading them. Use SERIAL_MCP_PLUGINS=name1,name2 to allow only specific plugins rather than all.
Writes affect real devices. A bad command sent to a serial device can trigger unintended behavior, disrupt other connected systems, or cause hardware damage (e.g., wiping flash, entering bootloader mode, triggering actuators). Consider what the agent can reach.
Use tool approval deliberately. When your MCP client prompts you to approve a tool call, consider whether you want to allow it once or always. "Always allow" is convenient but means the agent can repeat that action without further confirmation.
This software is provided as-is under the MIT License. You are responsible for what the agent does with your hardware.
This project is licensed under the MIT License — see LICENSE for details.
This project is built on top of the excellent pyserial library for cross-platform serial communication in Python.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.