MCP 服务器

mcp-agent-cost-optimizer

Enables precise financial analysis of AI agent costs, including token pricing, multi-step run estimates, model comparison, and ROI versus human labor, with deterministic decimal math.

README

mcp-agent-cost-optimizer

Exact LLM & AI Agent Cost Calculator + ROI Optimizer (Model Context Protocol server)

Give your AI agents (and yourself) high-precision financial math for token costs, multi-step agent runs, model comparisons, and ROI versus human labor. Pure computation — $0 running cost.

Why this exists

Entrepreneurs and AI builders waste time and money because:

Spreadsheets break on multi-step agent costs (retries, tool calls, context).
LLMs give approximate / wrong money numbers.
Hard to quickly compare "Claude vs GPT-4o vs Gemini for my agent fleet".

This MCP server solves it with deterministic, auditable decimal math.

Tools

Tool	What it does
`calculate_token_cost`	Exact $ for input + output tokens on any model (or custom prices)
`estimate_agent_run_cost`	Full agent workflow cost including steps + retry overhead
`compare_models`	Rank models by cost for the same workload
`calculate_agent_roi`	Monthly/annual savings & ROI of agent vs human
`list_supported_models`	See built-in pricing table

Quick start (local)

# Install
pip install -e .

# Run
python -m src.server
# or
mcp-agent-cost-optimizer

Add to Claude Desktop / Cursor / any MCP client:

{
  "mcpServers": {
    "agent-cost-optimizer": {
      "command": "python",
      "args": ["-m", "src.server"],
      "cwd": "/path/to/mcp-agent-cost-optimizer"
    }
  }
}

Example usage from an agent

"Estimate the cost of running a 5-step research agent 500 times a day on Claude 4 Sonnet with 15% retry rate, then compare vs GPT-4o-mini and tell me the ROI if a human analyst costs $60/hour and takes 20 minutes."

The agent will call the tools and return exact numbers with formulas.