Wraps the Tuning Engines governed AI runtime so Claude can fine-tune code models, monitor training jobs, inspect usage traces, and manage workflow policies through conversation. Exposes 60+ tools covering job creation with specialized agents like Cody for autocomplete and SIERA for bug fixes, cost estimation, model export to S3, credit balance checks, and approval workflows. Built for teams that want RBAC and audit trails around model training and inference routing without leaving their editor. The MCP layer intentionally blocks secret mutation and inference key creation, pushing those operations to the CLI or web UI. Most useful when you're already running Tuning Engines and want your AI assistant to kick off training runs or check job status without context switching.
claude mcp add --transport stdio io.github.cerebrixos-org-tuning-engines uvx tuning-engines