Logit Processor -- MCP Tool Specification

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

Overview

This document defines the MCP tools exposed by the logit_processor 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	logit_processor
Trust default	Safe
PAI bridge	src/codomyrmex/agents/pai/mcp/

Tool Reference

process_logits

Description: Apply sampling strategies to language model logits. Trust Level: Safe Category: analysis

Parameters:

Parameter	Type	Required	Default	Description
logits	list[float]	Yes	--	Raw logit values from language model
temperature	float	No	1.0	Scaling factor (above 1 is diverse, below 1 is focused, and 1 is unchanged)
top_k	int	No	0	Keep only top-k tokens (0=disabled)
top_p	float	No	1.0	Nucleus sampling threshold (1.0=disabled)
repetition_penalty	float	No	1.0	Penalize repeated tokens (1.0=disabled, >1.0=penalize)
previous_tokens	list[int] or None	No	None	Previously generated token IDs for repetition penalty
seed	int or None	No	None	Random seed for reproducibility

Returns: dict -- Dictionary with sampled_token, greedy_token, top5_tokens (list of mappings with id and prob), and entropy.

Example:

from codomyrmex.logit_processor.mcp_tools import process_logits

result = process_logits(
    logits=[2.0, 1.0, 0.5, -1.0, 3.0],
    temperature=0.8,
    top_k=3,
    seed=42,
)

Integration Notes

• Auto-discovered: Yes (via @mcp_tool in mcp_tools.py)
• Trust Gateway: All tools are safe -- no destructive operations
• PAI Phases: BUILD (token sampling), VERIFY (sampling distribution analysis)
• Dependencies: Requires numpy and internal processor module

Navigation

• README.md | AGENTS.md | SPEC.md