Free MCP Server for AI API Pricing — 42 Models in Claude Code & Cursor

Stop switching tabs to check API prices. Get live pricing for GPT-5.5, Claude Opus 4.8, Gemini, DeepSeek, and 38 more models directly inside your AI coding tools.

If you use Claude Code, Cursor, or Windsurf for development, you've probably asked your AI assistant "which model should I use?" or "how much will this cost?" only to get outdated or hallucinated pricing data.

That's the problem the APIpulse MCP server solves. It's a free Model Context Protocol server that gives your AI tools live access to real pricing data for 42 models across 10 providers.

Install in 10 Seconds

One command. No npm install. No API key. No signup.

claude mcp add apipulse --transport http https://getapipulse.com/api/mcp

For Cursor, add this to your MCP settings (.cursor/mcp.json):

{
  "mcpServers": {
    "apipulse": {
      "url": "https://getapipulse.com/api/mcp"
    }
  }
}

For Windsurf or any other MCP client, use the endpoint: https://getapipulse.com/api/mcp

5 Tools, 42 Models, 10 Providers

Once installed, your AI assistant can query live pricing data with these tools:

Supported Providers

Example Prompts

After installing, try asking your AI assistant:

• "What's the cheapest model for code generation?" — Uses find_cheapest to compare budget-tier models
• "Compare Claude Sonnet 4.6 vs GPT-5 pricing" — Uses compare_models for side-by-side breakdown
• "How much would 100K input + 50K output tokens cost on Gemini 3 Flash?" — Uses calculate_cost
• "Is Claude 4 Opus deprecated? What should I use instead?" — Uses get_model_details with replacement info
• "What's the cheapest model with a 1M context window?" — Uses find_cheapest with context filter

Why an MCP Server for Pricing?

AI model pricing changes fast. In the last 6 months alone:

• Claude 4 was shut down and replaced by Claude Opus 4.8 and Sonnet 4.6
• DeepSeek V4 Flash launched at $0.14/$0.28 per 1M tokens — 97% cheaper than GPT-5.5
• Google shipped Gemini 3.5 Flash with 1M context at budget pricing
• OpenAI released GPT-5 mini at $0.25/$2 — making "expensive" models obsolete overnight

Hardcoded pricing tables go stale. The APIpulse MCP server always returns current data because it reads from a live API, not a static file.

How It Works

The APIpulse MCP server implements the MCP Streamable HTTP transport (2025-03-26 spec). It's a standard JSON-RPC endpoint — no special protocol libraries needed.

When your AI assistant uses one of the 5 tools, the MCP server returns structured JSON with current pricing data. The data covers:

• 42 models — from budget ($0.075/1M input) to premium ($30/1M input)
• 10 providers — OpenAI, Anthropic, Google, DeepSeek, Mistral, Cohere, Meta, Moonshot, xAI, AI21
• 3 tiers — budget, mid, premium
• Deprecation tracking — know which models are sunsetting and what replaces them

Open Data, Open Source

The pricing data is also available as a static JSON API under a CC-BY-4.0 license:

curl https://getapipulse.com/data/pricing.json

Use it in your own tools, dashboards, or scripts. An OpenAPI spec is also available at /data/pricing-openapi.json.