SPICEBridge

AI-powered circuit design through simulation — an MCP server that gives language models direct access to SPICE circuit simulation via ngspice. Describe a circuit in plain English and let the AI handle netlist generation, simulation, measurement, and spec verification.

Features

• 28 tools covering the full circuit design workflow
• 11 built-in templates with automatic component value calculation (E24 series)
• Simulation: AC sweep, transient, DC operating point
• Measurement: bandwidth, gain, DC levels, transient metrics, power
• Monte Carlo & worst-case analysis under component tolerances
• Multi-stage composition — connect circuit stages with automatic port mapping
• Model wizard — generate SPICE .lib models from datasheet parameters
• KiCad export — output .kicad_sch schematics
• Web viewer — interactive schematic viewer in the browser
• Cloud setup wizard — one-command deployment with Cloudflare tunnels (spicebridge setup-cloud)
• Spec verification — compare results against design targets

Install

pip install spicebridge

Requirements

• Python 3.10+
• ngspice installed and on PATH

Quick Start

Use the public server (no install needed)

The fastest way to get started — no Python, no ngspice, nothing to install. Just add this to your MCP client config and start designing circuits:

Claude Desktop / Claude.ai — add to your MCP settings:

{
  "mcpServers": {
    "SPICEBridge": {
      "url": "https://spicebridge.clanker-lover.work/mcp"
    }
  }
}

Claude Code — run this from any project directory:

claude mcp add --transport http SPICEBridge https://spicebridge.clanker-lover.work/mcp

That's it. You're ready to go.

Self-hosted options

If you'd rather run your own instance:

Local (Claude Code / stdio)

Add to your .mcp.json:

{
  "mcpServers": {
    "spicebridge": {
      "command": "spicebridge"
    }
  }
}

Cloud (Streamable HTTP)

spicebridge --transport streamable-http --port 8000

Cloud Setup Wizard

One command to go from a local install to a public MCP endpoint:

spicebridge setup-cloud          # interactive (named tunnel, permanent URL)
spicebridge setup-cloud --quick  # quick tunnel (temporary URL, no account needed)

The wizard handles the full deployment pipeline:

1. Installs cloudflared automatically (macOS via Homebrew, Linux via APT) if not found
2. Authenticates with Cloudflare (browser-based OAuth, named tunnel mode only)
3. Creates or reuses a tunnel — prompts to pick an existing one or make a new one
4. Configures DNS routing for your custom domain (named tunnel mode)
5. Generates an API key for authentication
6. Starts the SPICEBridge server and tunnel together
7. Prints connection info with a ready-to-paste JSON config for Claude.ai

Quick tunnels give you a temporary trycloudflare.com URL instantly — no Cloudflare account required. Named tunnels give you a permanent custom domain (e.g. spicebridge.example.com).

Additional options:

spicebridge setup-cloud --domain mcp.example.com  # specify custom domain
spicebridge setup-cloud --port 9000                # custom port
spicebridge setup-cloud --no-install               # skip cloudflared installation

Example

1. load_template("rc_lowpass_1st", specs={"f_3dB_hz": 1000})
   -> netlist with R=1.6k, C=100nF, circuit_id: "a1b2c3d4"

2. run_ac_analysis(circuit_id, start_freq=1, stop_freq=1e6)
   -> frequency response data

3. measure_bandwidth(circuit_id)
   -> f_3dB_hz: 995

4. compare_specs(circuit_id, specs={"f_3dB_hz": {"target": 1000, "tolerance_pct": 5}})
   -> PASS

5. draw_schematic(circuit_id)
   -> schematic SVG

Tools

Create & Configure

Simulate

Measure

Evaluate & Export

Composition & Ports

Advanced Analysis

Model Management

Development

git clone https://github.com/clanker-lover/spicebridge.git
cd spicebridge
pip install -e ".[dev]"
pytest

License

GPL-3.0-or-later