Quantization -- MCP Tool Specification

Version: v1.1.9 | Status: Active | Last Updated: March 2026

Overview

This document defines the MCP tools exposed by the quantization module. These tools are auto-discovered by the PAI MCP bridge via the @mcp_tool decorator in mcp_tools.py and surfaced as part of the ~303 dynamic tools available to Claude.

Auto-Discovery

Property	Value
Discovery method	@mcp_tool decorator scan
Namespace	quantization
Trust default	Safe
PAI bridge	src/codomyrmex/agents/pai/mcp/

Tool Reference

quantize_tensor

Description: Quantize a list of float values to int8 or fp4. Trust Level: Safe Category: data-mutation

Parameters:

Parameter	Type	Required	Default	Description
values	list[float]	Yes	--	List of float32 values to quantize
method	str	No	"int8"	Quantization method -- "int8" or "fp4"
scheme	str	No	"asymmetric"	For int8: "symmetric" or "asymmetric" (ignored for fp4)

Returns: dict -- Dictionary with method, quantized values, scale, reconstruction error metrics, and status.

Example:

from codomyrmex.quantization.mcp_tools import quantize_tensor

result = quantize_tensor(
    values=[1.5, -0.3, 2.1, 0.0, -1.7],
    method="int8",
    scheme="symmetric",
)

---

quantization_benchmark

Description: Benchmark int8 vs fp4 quantization on random data. Trust Level: Safe Category: analysis

Parameters:

Parameter	Type	Required	Default	Description
size	int	No	1000	Number of float32 values to benchmark

Returns: dict -- Comparison of int8 symmetric, int8 asymmetric, and fp4 error metrics with timing.

Example:

from codomyrmex.quantization.mcp_tools import quantization_benchmark

result = quantization_benchmark(size=5000)

Integration Notes

• Auto-discovered: Yes (via @mcp_tool in mcp_tools.py)
• Trust Gateway: All tools are safe -- no destructive operations
• PAI Phases: BUILD (model quantization), VERIFY (error analysis benchmarking)
• Dependencies: Requires numpy and internal int8, fp4, utils modules

Navigation

• README.md | AGENTS.md | SPEC.md