Skip to content

MCP Tool Search Budget Simulator

Simulate context-window usage for full MCP tool injection versus search-first retrieval strategies

MCP Tool Search Budget Simulator

{
  "input": {
    "contextWindow": 200000,
    "toolCount": 320,
    "avgToolSchemaTokens": 180,
    "queryOverheadTokens": 1400,
    "retrievedTools": 14
  },
  "fullInjection": 59000,
  "searchInjection": 3920,
  "savings": 55080,
  "savingsPct": 93.4,
  "status": "healthy",
  "recommendedRetrieved": 320,
  "notes": [
    "Keep full catalog out of default prompt when possible.",
    "Use tool search only after intent is classified.",
    "Cap retrieved tools to avoid latency and context churn."
  ]
}

Simulation complete

Updated . Provided as is. Check the output before you rely on it in production.

How to use MCP Tool Search Budget Simulator

  1. 1

    Provide workload sizing inputs

    Enter context window size, total tool count, average schema token size, query overhead, and retrieved tool count in JSON format.

  2. 2

    Run simulation for both injection strategies

    Execute Run Simulation to compute full injection and search-first token footprints side by side.

  3. 3

    Inspect savings and overflow status

    Review savings percentage, utilization, and overflow indicators to decide whether your current settings are safe for production traffic.

  4. 4

    Tune retrieved tools for safe headroom

    Use the recommendation output to adjust retrieval depth and re-run until you maintain acceptable context-window margin.

Questions and answers

What is MCP Tool Search Budget Simulator?
Every MCP tool schema loaded into a prompt uses context tokens, so large tool catalogs crowd out the conversation. This simulator compares injecting every schema with retrieving only the top matches per query, and reports tokens used, savings, status and a recommended retrieval count for your context window.
Which inputs are required?
Provide context window size, total tool count, average schema token size, retrieval overhead, and retrieved tool count. The simulator then calculates utilization, savings, and overflow risk.
What does overflow status mean?
Overflow means injecting every tool schema plus the query overhead exceeds your context window; warning means it fills more than 80% of it. Reduce tool count or schema size, or switch to search-first retrieval to restore headroom.
Does MCP Tool Search Budget Simulator store or send my data?
No. All calculations run locally in the browser and your planning inputs remain private.
For AI agents: how to call this tool

Machine-readable contract, endpoints and examples. Humans can ignore this section.

Best Path For Builders

Browser workflow

Runs instantly in the browser with private local processing and copy/export-ready output.

Browser Workflow

This tool is optimized for instant in-browser execution with local data handling. Run it here and copy/export the output directly.

/mcp-tool-search-budget-simulator/

For automation planning, fetch the canonical contract at /api/tool/mcp-tool-search-budget-simulator.json.