Agent Cost Hub
Precios de API de LLM y estimaciones de costos de agentes de IA, verificados a diario. Herramientas de solo lectura: list_models, get_model_price, estimate_cost, compare_models, cheapest_models_for. 79 modelos, gratis, sin autenticación. HTTP transmisible: https://agentcosthub.com/mcp
Servidor MCP alojado
npx add-mcp 'https://agentcosthub.com/mcp'Se instala en Claude Code, Codex, Cursor y más
Documentación
Use our prices inside your AI
Ask Claude or Cursor "what would this agent cost per month?" and get an answer from the same daily-checked prices as this site. It's free, there's no sign-up and no API key.
Connector URL: https://agentcosthub.com/mcp
What is MCP?
MCP (Model Context Protocol) is a standard way to plug extra tools into an AI assistant. Once you add our connector, your AI can look up model prices and calculate agent costs for you, instead of guessing from old training data. Every answer includes the date prices were last verified and a link to the matching page here.
Set it up in Claude.ai (browser)
- Open claude.ai and go to Customize → Connectors.
- Click + Add, then Add custom connector.
- Give it a name (e.g. "Agent Cost Hub") and paste the URL
https://agentcosthub.com/mcp. Click Continue. - For authentication, keep No sign-in and leave headers empty. Click Add.
- Start a new chat and ask one of the example questions below.
Claude's free plan allows one custom connector.
Set it up in Claude Desktop
Easiest: open Settings (Ctrl+, or Cmd+,) → Connectors → Add → Add custom connector and follow the same steps as above. Connectors belong to your Claude account, so one added on claude.ai also appears in the desktop app. Alternative (runs a small bridge on your computer):
- Install Node.js if you don't have it.
- In Claude Desktop open Settings → Developer → Edit Config.
- Paste this into
claude_desktop_config.json(merge it if you already have servers), save, and restart Claude Desktop:
{ "mcpServers": { "agentcosthub": { "command": "npx", "args": ["-y", "mcp-remote", "https://agentcosthub.com/mcp"] } } }
Set it up in Cursor
- Open or create
~/.cursor/mcp.json(all projects) or.cursor/mcp.jsonin one project. - Paste this and save:
{ "mcpServers": { "agentcosthub": { "url": "https://agentcosthub.com/mcp" } } }
Then open Customize in Cursor's sidebar and check that the server is listed and switched on. Steps checked 2 Nov 2026 against Claude's and Cursor's help pages.
Example questions
- What would a coding agent on Grok 4.7 cost per month?
- Compare Claude Sonnet 5.5 and GPT-6.1 Sol for a chatbot with 500 chats a day.
- What are the 5 cheapest models for a RAG app?
- How much does DeepSeek V4.1 Flash cost per million tokens?
Tools your AI gets
- list_models: all 79 models with prices per 1M tokens.
- get_model_price: one model's input, output and cached prices.
- estimate_cost: monthly cost for one model. Name a ready-made workload (chatbot, rag, coding_agent or research_agent) or give your own tokens per call, calls per day, steps and caching.
- compare_models: 2 to 5 models side by side.
- cheapest_models_for: cheapest models for a chatbot, RAG, coding agent or research agent.
All tools are read-only and use the same formula as our methodology.
Privacy
- The connector only receives what your AI sends it: a model name and a few numbers (tokens, steps, tasks per day). It never sees your prompts, code, files or chat.
- It's read-only: it can't change anything in your AI app or accounts.
- Your AI app keeps your chat history under its own privacy settings, as usual.
No account, no API key, no cookies. We don't store your questions or the numbers your AI sends; each request is answered and forgotten. Our host, Cloudflare, keeps standard request logs for security. See our privacy policy.
Questions? [email protected]