LLM Advisor

llm-advisor-mcp

English | 日本語

Give your AI assistant real-time LLM/VLM knowledge. Pricing, benchmarks, and recommendations — updated every hour, not every training cycle.

LLMs have knowledge cutoffs. Ask Claude "what's the best coding model right now?" and it cannot answer with current data. This MCP server fixes that by feeding live model intelligence directly into your AI assistant's context window.

• Zero config — No API keys, no registration. One command to install.
• Low token — Compact Markdown tables (~300 tokens), not raw JSON (~3,000 tokens). Your context window matters.
• 5 benchmark sources — SWE-bench, LM Arena Elo, OpenCompass VLM, Aider Polyglot, and OpenRouter pricing merged into one unified view.

---

Use Cases

• "What's the best coding model right now?" — list_top_models with category coding
• "Compare Claude vs GPT vs Gemini" — compare_models with side-by-side table
• "Find a cheap model with 1M context" — recommend_model with budget constraints
• "What benchmarks does model X have?" — get_model_info with percentile ranks

---

Quick Start

Claude Code

claude mcp add llm-advisor -- npx -y llm-advisor-mcp

Claude Code (Windows)

claude mcp add llm-advisor -- cmd /c npx -y llm-advisor-mcp

Claude Desktop / Cursor / Windsurf

Add to your MCP configuration file:

{

"mcpServers": { "llm-advisor": { "command": "npx", "args": ["-y", "llm-advisor-mcp"] } } }

That is it. No API keys, no .env files.

Compatible Clients

---

Tools

get_model_info

Detailed specs for a specific model: pricing, benchmarks, percentile ranks, capabilities, and a ready-to-use API code example.

Parameters

Example output

## anthropic/claude-sonnet-4

Provider: anthropic | Modality: text+image→text | Released: 2025-06-25

Pricing

Benchmarks

Percentile Ranks

Capabilities: Tools, Reasoning, Vision

API Example (openai_sdk)

from openai import OpenAI

client = OpenAI( base_url="https://openrouter.ai/api/v1", api_key=" ", ) response = client.chat.completions.create( model="anthropic/claude-sonnet-4", messages=[{"role": "user", "content": "Hello"}], )

```

---

list_top_models

Top-ranked models for a category. Includes release dates for freshness awareness.

Parameters

Example output

## Top 5: coding

---

compare_models

Side-by-side comparison for 2-5 models. Best values are bolded automatically. Includes a Released row so you can spot outdated models at a glance.

Parameters

Example output

## Model Comparison (3 models)

---

recommend_model

Personalized top-3 recommendations. Scores combine weighted benchmarks, pricing, capability bonuses, and a freshness bonus (+3 points for models released within 3 months, +1 within 6 months).

Parameters

Example output

## Recommended for: coding

1. anthropic/claude-sonnet-4 (score: 78)

Input: $3.00/1M | Output: $15.00/1M | Context: 200K | Released: 2025-06-25 Benchmarks: SWE-bench: 76.8%, Aider: 72.1%, Arena: 1467 Strengths: reasoning, tools, vision

2. google/gemini-2.5-flash (score: 74)

Input: $0.15/1M | Output: $0.60/1M | Context: 1M | Released: 2025-05-20 Benchmarks: SWE-bench: 62.9%, Arena: 1445 Strengths: tools, vision, 1M+ context

3. openai/o4-mini (score: 71)

Input: $1.10/1M | Output: $4.40/1M | Context: 200K | Released: 2025-04-16 Benchmarks: SWE-bench: 73.6%, Arena: 1430 Strengths: reasoning, tools

---

Data Sources

All data is fetched in real time from free, public APIs. No authentication required.

---

Context Cost

MCP tool definitions and responses consume your LLM's context window. This server is designed to be lean:

For comparison, most MCP servers that return raw JSON consume 3,000-10,000 tokens per response. Every response from llm-advisor-mcp is pre-formatted Markdown, keeping context costs roughly 10x lower.

---

Architecture

┌──────────────────────────────────────────────┐

│ MCP Client (Claude, etc.) │ └──────────┬───────────────────────────────────┘ │ stdio (JSON-RPC) ┌──────────▼───────────────────────────────────┐ │ llm-advisor-mcp server │ │ │ │ ┌─────────┐ ┌───────────┐ ┌────────────┐ │ │ │ Tools │ │ Registry │ │ Cache │ │ │ │ (4 tools)│──│ (unified) │──│ (in-memory)│ │ │ └─────────┘ └───────────┘ └────────────┘ │ │ │ │ │ ┌────────────┼────────────┐ │ │ ▼ ▼ ▼ │ │ ┌──────────┐ ┌──────────┐ ┌──────────┐ │ │ │Normalizer│ │Percentile│ │ Fetchers │ │ │ │(slug map)│ │ (5 cats) │ │(5 sources│ │ │ └──────────┘ └──────────┘ └──────────┘ │ └──────────────────────────────────────────────┘ │ │ │ OpenRouter SWE-bench Arena / VLM / Aider

• TypeScript + ESM — Single entry point, tsup build
• In-memory cache — TTL-based (1h pricing, 6h benchmarks), stale-while-revalidate
• Cross-source normalization — Maps inconsistent model names (e.g. Claude 3.5 Sonnet vs anthropic/claude-3.5-sonnet) to canonical IDs
• Percentile computation — Ranks across 5 categories (coding, math, general, vision, cost efficiency)
• Freshness scoring — Recommendation algorithm gives a bonus to recently released models (+3 for <=3mo, +1 for <=6mo)
• Zero runtime deps beyond @modelcontextprotocol/sdk and zod

---

Roadmap

---

Development

git clone https://github.com/Daichi-Kudo/llm-advisor-mcp.git

cd llm-advisor-mcp npm install npm run build # Build with tsup npm run dev # Run with tsx (hot reload) npm test # Run 51 unit tests (vitest) npm run test:watch # Watch mode

Project structure

src/

index.ts # Server entry point types.ts # Shared type definitions tools/ model-info.ts # get_model_info tool list-top.ts # list_top_models tool compare.ts # compare_mo