LLM Prices CN MCP Server - LLM API Pricing Dataset¶

Repo: szp2005/llm-prices-cn Pricing: Free, hosted Category: Finance / Cost Optimization

Overview¶

LLM Prices CN is a daily-verified dataset of LLM API pricing covering 44+ models across Chinese and global providers. It includes a hosted MCP server for live price queries and token cost estimation - critical for business operators managing AI agent infrastructure costs.

Why Business Operators Need This¶

Cost Visibility - With 44+ LLM models available, pricing changes frequently. LLM Prices CN verifies prices daily so your cost estimates are always current.
Provider Comparison - Compare input/output token pricing across OpenAI, Anthropic, DeepSeek, Moonshot, Zhipu, Qwen, and more in one query.
Budget Planning - Estimate monthly LLM costs based on expected token volume before committing to a provider.
China + Global Coverage - Unique coverage of Chinese LLM providers (DeepSeek, Moonshot, Zhipu, Qwen) alongside global providers - essential for operators with multi-market needs.

Installation¶

Hosted (Remote)¶

The server is hosted and requires no local installation:

{
  "mcpServers": {
    "llm-prices": {
      "url": "https://llm-prices-cn.example.com/mcp",
      "transport": "streamable-http"
    }
  }
}

Local¶

{
  "mcpServers": {
    "llm-prices": {
      "command": "python",
      "args": ["-m", "llm_prices_cn"]
    }
  }
}

Key Tools¶

Tool	Function
`get_all_prices`	Retrieve current pricing for all 44+ tracked models
`get_model_price`	Get pricing for a specific model (e.g., `gpt-4o`, `claude-sonnet-4`)
`estimate_cost`	Calculate estimated cost for given token volumes
`compare_providers`	Side-by-side pricing comparison for equivalent capability tiers
`price_history`	Track pricing changes over time for a model
`get_lowest_cost`	Find the cheapest model meeting your capability requirements

Covered Providers & Models (44+)¶

Provider	Models Tracked	Region
OpenAI	GPT-4o, GPT-4o-mini, o3, o4-mini	Global
Anthropic	Claude Opus 4, Sonnet 4, Haiku 3.5	Global
DeepSeek	V3, R1, V4-Pro	CN/Global
Moonshot	Kimi series	CN
Zhipu AI	GLM-4 series	CN
Qwen	Qwen-Max, Qwen-Plus	CN
Google	Gemini 2.5 Pro, Flash	Global
xAI	Grok 3	Global
SiliconFlow	Various open-source	CN
Groq	Llama, Mixtral	Global
Together AI	Various	Global

Pricing verified daily. Last verification date included in every response.

Common Queries¶

Cost Estimation¶

"Estimate my monthly LLM costs: 500K input tokens/day, 100K output tokens/day. Compare GPT-4o vs Claude Sonnet 4."
"What's the cheapest model that can handle 128K context for my document processing pipeline?"
"If I switch from Claude Opus 4 to DeepSeek V4-Pro, how much will I save monthly?"

Provider Comparison¶

"Compare input/output pricing for all Chinese LLM providers. I need 1M tokens/day."
"Show me the pricing history for GPT-4o over the last 3 months - has it changed?"
"Which provider has the best price/performance ratio for coding tasks?"

Budget Planning¶

"I have a $500/month LLM budget. How many tokens can I process with each provider?"
"Alert me if any provider changes prices by more than 5%."
"What's the total cost difference between using OpenAI exclusively vs a multi-provider setup?"

Best Practices¶

Check prices weekly - LLM pricing changes frequently, especially among Chinese providers competing aggressively
Use estimate_cost before large jobs - A 10M token batch job can cost $0.50 or $15.00 depending on the model
Compare tiers fairly - Match capability tiers (e.g., GPT-4o vs Claude Sonnet 4, not GPT-4o vs Haiku)
Factor in quality - Lowest cost isn't always best. Cross-reference with your quality requirements
Monitor price_history - Track trends to spot providers getting cheaper (or more expensive) over time

Integration with CorpusIQ¶

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects

Limitations¶

New server (July 2026) - data freshness guarantees still being established
China-focused dataset - some global providers may have less frequent updates
Token cost estimation assumes standard tokenization - actual costs vary with prompt structure
Hosted endpoint URL may change - verify the current endpoint on the GitHub repo

For Operators: Integration with CorpusIQ¶

CorpusIQ QuickBooks connector + LLM Prices CN = Track actual LLM spend against estimated costs
CorpusIQ Stripe connector + LLM Prices CN = Monitor API billing against provider pricing
CorpusIQ cross-source analysis + LLM Prices CN = Build LLM cost dashboards with actual vs budget tracking