MCP Server

Genalyzer provides a Model Context Protocol (MCP) server that exposes spectral and code-density analysis tools to AI assistants and other MCP clients. Tools cover all five genalyzer analysis domains — Fourier, histogram, DNL, INL, and time-domain waveform — plus generators and a quantizer for end-to-end simulate-and-verify workflows.

Installation

pip install genalyzer[mcp]

The MCP server also requires the native libgenalyzer library. See Installation for build instructions.

Running the server

genalyzer-mcp

Or as a module:

python -m genalyzer.mcp.server

Claude Desktop configuration

{
  "mcpServers": {
    "genalyzer": {
      "command": "genalyzer-mcp"
    }
  }
}

Tool inventory

All tools read .npy or .csv inputs (auto-detected by extension). Every analyze_* tool accepts a plot: bool = False flag; set to True to render an annotated PNG next to the input.

Tool	Layer	Purpose
generate_test_tone	Generator	Complex sinusoid
generate_real_tone	Generator	Real sinusoid
generate_ramp	Generator	Linear ramp
generate_gaussian_noise	Generator	AWGN
quantize	Primitive	Apply N-bit quantization
compute_fft	Primitive	FFT of time-domain data
get_fa_metrics	Primitive	Metrics from pre-computed FFT
compute_histogram	Primitive	Code-count histogram
compute_dnl	Primitive	DNL from histogram
compute_inl	Primitive	INL from DNL
compute_waveform_stats	Primitive	Time-domain stats
analyze_spectrum	All-in-one	Load → FFT → metrics (+ optional PNG)
analyze_histogram	All-in-one	Load → histogram → metrics (+ optional PNG)
analyze_dnl	All-in-one	Load → DNL → metrics (+ optional PNG)
analyze_inl	All-in-one	Load → INL → metrics (+ optional PNG)
analyze_waveform	All-in-one	Time-domain analysis (+ optional PNG)

Pipeline

See also

• Workflows
• API Reference