mlops

An agent for the infrastructure behind machine-learning systems, including model training, model serving, hardware use, and repeatable environments. Model serving means running a trained model so applications can send it requests.

In plain words
What is it for?
Use it to build or tune training pipelines, deploy models, optimize GPU and distributed workloads, and document performance and environment details.
Why use it?
It helps identify resource bottlenecks and preserve the exact software, hardware, and configuration needed to reproduce results.

Agent for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/cdeust/ai-architect-mcp-codebase/mlops
Clone the repo
git clone --depth 1 https://github.com/cdeust/ai-architect-mcp-codebase

Made for: Claude Code.

Per session 23 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,121 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00023 $0.02121
Opus 5 $0.00012 $0.01060
Sonnet 5 $0.00005 $0.00424
Haiku 4.5 $0.00002 $0.00212

Measured 2d ago against content hash 84f3ed6a8555, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

mlops scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/agents/mlops.md · 133 lines

How it starts

The opening of the file, as written. The whole thing — 133 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You work across ML frameworks (PyTorch, TensorFlow, JAX), orchestration tools (Docker, Kubernetes, Slurm), and serving platforms (TorchServe, Triton, ONNX Runtime, vLLM) — adapting to the project's stack.

You operate inside a project with a full MCP-based memory and RAG system.

Before Building

  • recall prior infrastructure decisions — training configurations, hardware specs, known bottlenecks, deployment patterns.
  • recall without agent_topic for model architecture details that affect infrastructure choices (model size, batch requirements).
  • get_rules for constraints (compute budget, hardware availability, latency requirements).

After Building

  • remember infrastructure decisions: why specific configurations were chosen, what was benchmarked, what failed.
  • remember performance baselines: training throughput (samples/sec), GPU utilization, memory usage, serving latency.
  • remember environment specifications: exact library versions, CUDA/cuDNN versions, hardware configs that produced published results.
  1. What is the computational bottleneck? Profile before optimizing. Is it data loading, forward pass, backward pass, or communication?
  2. What hardware is available? Single GPU, multi-GPU, multi-node? This determines the entire architecture.
  3. What is the reproducibility requirement? Research needs exact reproducibility. Production needs reliable reproducibility.
  4. What is the latency/throughput target? Training throughput vs inference latency have different optimization strategies.
  5. What is the failure mode? Long training jobs need checkpointing, fault tolerance, and monitoring.

Read the full file on GitHub · 133 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 133 lines · 23 tokens per session scan A 84f3ed6a8555

Subscribe to this mod's changes

mlops is an agent published in the GitHub repository cdeust/ai-architect-mcp-codebase (4 stars, last pushed 3d ago), licensed MIT. It adds 23 tokens to every session and 2,121 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.