Distributed Training -- MCP Tool Specification
Version: v1.1.9 | Status: Active | Last Updated: March 2026
Overview
This document defines the MCP tools exposed by the distributed_training 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 | distributed_training |
| Trust default | Safe |
| PAI bridge | src/codomyrmex/agents/pai/mcp/ |
Tool Reference
fsdp_simulate_step
Description: Simulate one FSDP distributed training step. Trust Level: Safe Category: analysis
Parameters:
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
param_size | int | No | 1024 | Number of model parameters |
world_size | int | No | 4 | Number of simulated GPU devices |
learning_rate | float | No | 0.01 | SGD learning rate |
seed | int | No | 42 | Random seed for reproducibility |
Returns: dict -- Dictionary with shard_sizes, param_norm_before, param_norm_after, grad_norm, and status.
Example:
from codomyrmex.distributed_training.mcp_tools import fsdp_simulate_step
result = fsdp_simulate_step(param_size=2048, world_size=8, learning_rate=0.001)
Integration Notes
- Auto-discovered: Yes (via
@mcp_toolinmcp_tools.py) - Trust Gateway: All tools are safe -- no destructive operations
- PAI Phases: BUILD (training simulation), VERIFY (performance validation)
- Dependencies: Requires
numpyand internalfsdpmodule