Log inSubmit server

MCP Token Cost Calculator

Every MCP server you connect sends its tool definitions to the model on every turn. Estimate how much of your context window that's actually costing you.

  • Free
  • No sign-in
  • Runs in your browser

1 Your tools

Paste the raw tools/list response from your server (or a bare array of tool schemas). The example below is a small GitHub-style server.

tools/list response

2 Context cost

158tokens on every turn
  • 128K GPT-4 class0.12%
  • 200K Claude0.08%
  • 1M Claude long context0.02%

Per tool (2)

  • search_issues92
  • create_issue66

Counts use a GPT-4-class tokenizer as an approximation — exact counts vary by model.

Why MCP servers have a token cost

When an MCP client connects to a server, it asks for that server's list of tools — each with a name, a description, and a JSON Schema describing its parameters — and includes all of that in every request sent to the model, whether or not the model ends up calling any of those tools that turn. Connect enough servers, especially ones with many tools or verbose parameter schemas, and you can burn a meaningful slice of your context window before you've typed a single message.

Frequently asked questions

A typical MCP tool schema consumes between 150 and 500 tokens per tool. A server exposing 10 tools with detailed parameter descriptions can use 1,500 to 5,000 tokens on every prompt turn.

When an MCP client connects to a server, it prepends all registered tool definitions into the system context on every conversation message so the AI model knows available tools.

To reduce token cost: disable unused MCP servers when not in active use, shorten overly verbose tool descriptions, and consolidate similar parameter objects.