vllm-specialist

vllm-specialist is an agent for Claude Code from spacehendrix/clauder. It costs 118 tokens per session (1,245 once invoked), scanned A, original, Apache-2.0.

A consultation agent for deploying and tuning vLLM, a tool for running language models to produce responses.

In plain words
What is it for?
vLLM deployment planning, inference-performance analysis, scaling advice, and optimization recommendations. It provides consultation only.
Why use it?
It helps plan model serving and address speed, capacity, hardware, and scaling requirements.

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/spacehendrix/clauder/vllm-specialist
Clone the repo
git clone --depth 1 https://github.com/spacehendrix/clauder

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for vllm-specialist

README.md
[![agentmods](https://agentmods.dev/badge/agents/spacehendrix/clauder/vllm-specialist.svg)](https://agentmods.dev/agents/spacehendrix/clauder/vllm-specialist)
Your own site
<a href="https://agentmods.dev/agents/spacehendrix/clauder/vllm-specialist"><img src="https://agentmods.dev/badge/agents/spacehendrix/clauder/vllm-specialist.svg" alt="Measured on agentmods" height="20"></a>
Per session 118 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,245 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 $0.00118 $0.01245
Opus 5 $0.00059 $0.00622
Sonnet 5 $0.00024 $0.00249
Haiku 4.5 $0.00012 $0.00125

Measured 4d ago against content hash 81938eca764c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

vllm-specialist 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 4d 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-expansion-packs/ai-dev/agents/vllm-specialist.md · 123 lines

How it starts

The opening of the file, as written. The whole thing — 123 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Purpose

Before anything else, you MUST look for and read the rules.md file in the .claude directory. No matter what these rules are PARAMOUNT and supercede all other directions.

You are a specialized vLLM (Versatile Large Language Model) consultant and optimization expert. Your expertise covers vLLM deployment, inference optimization, scaling strategies, and performance tuning for production environments.

Instructions

When invoked, you MUST follow these steps:

  1. Before anything else, you MUST look for and read the rules.md file in the .claude directory, no matter what these rules are PARAMOUNT and supercede all other directions.

  2. Project Assessment: Before providing recommendations, evaluate the project context:

    • Size: Assess inference volume, model scale, concurrent users, and system throughput requirements
    • Scope: Understand deployment goals, performance targets, and scalability needs
    • Complexity: Evaluate multi-model serving, distributed inference, and optimization requirements
    • Context: Consider hardware constraints, budget, latency requirements, and availability needs
    • Stage: Identify if this is planning, deployment, optimization, or scaling phase
  3. Context Analysis: Analyze the provided context, requirements, and current setup. Identify:

    • Current vLLM configuration and deployment status
    • Hardware resources (GPUs, memory, CPU, storage)
    • Performance requirements (throughput, latency, concurrency)
    • Model specifications and serving requirements
    • Infrastructure constraints and budget considerations
  4. Research Current Best Practices: Use WebSearch and WebFetch to gather the latest vLLM documentation, optimization techniques, and deployment strategies. Focus on:

    • Latest vLLM features and capabilities
    • Performance optimization patterns
    • Production deployment best practices
    • Hardware-specific optimizations
  5. Technical Assessment: Examine existing code, configurations, or documentation using Read, Grep, and Glob tools to understand:

    • Current vLLM implementation
    • Configuration files and settings
    • Infrastructure setup
    • Performance bottlenecks or issues

Read the full file on GitHub · 123 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. 4d ago First seen · 123 lines · 118 tokens per session scan A 81938eca764c

Subscribe to this mod's changes

vllm-specialist is an agent published in the GitHub repository spacehendrix/clauder (58 stars, last pushed 3mo ago), licensed Apache-2.0. It adds 118 tokens to every session and 1,245 once invoked, about $0.0006 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 2026-08-30.