mcpcost

io.github.zhaoyue722/llm-usage-mcp

LLM Usage

Is the LLM Usage MCP server down? No. It answered initialize and tools/list in the probe of 2026-10-31.

Local-first, multi-provider tool that captures LLM API spend and exposes it to coding agents via MCP

alive probed 2026-11-15 · unchanged since 2026-10-31

7tools
3.9ktokens in Claude Opus 5.5
1.9 %of Claude's 200k window
1.5×the median server
4.5 sto initialize
7 msto list tools
Run it
uvx llm-usage-mcp@0.1.3 (stdio, probed in a container with the network off)
Host
pypi.org
Reports itself as
llm-usage 1.30.0
Capabilities
experimental, prompts, resources, tools
Instructions
none
Of a 1M window
0.4 % of a 1M context window
Count
3.9k tokens in Claude Opus 5.5, exact, measured 2026-10-17. With the Claude Code prefix mcp__…__ on each tool name: 4.0k. The o200k_base estimate is 2.2k. Counted again when the tool definitions change. Method.
Repository
https://github.com/zhaoyue722/llm-usage-mcp
Website
https://github.com/zhaoyue722/llm-usage-mcp
Registry
active, updated 2026-08-03 · registry entry

Most expensive tools

The Claude column scales the o200k estimate of each tool by this server's ratio of exact to estimate, 1.74×.

Tool≈ Claude tokenso200k estimateShare of this server
recommend_provider92353124 %
query_spend84748722 %
record_usage60534816 %
compare_providers59634315 %
usage_summary46126512 %

Compare this server with others.

More servers

Other registry entries on pypi.org, the ones next to this name; all of them.

Servers that mention usage.

Badge for the README

mcpcost: alive

The result of the latest probe, the tool count and the token cost. The feed has an entry for every probe that changed them.

Atom feed

A mail when this server changes

Stops answering, comes back, tool set or cost changes. One confirmation mail first, a stop link in every mail.

For the publisher