⚡ Instant Tools

MCP Tool Checker

Score MCP, OpenAI, and Anthropic tool schemas for agent readiness—descriptions, naming, safety, and context size. Free, no signup.

Privacy: Your tool manifests and OpenAPI specs never leave this browser. Scoring is 100% client-side.

1. Paste tools JSON or OpenAPI

How to Use

  1. 1Paste MCP tools/list JSON, OpenAI/Anthropic tool defs, or upload an OpenAPI file.
  2. 2Run Check tools to get a 0–100 readiness score across five dimensions.
  3. 3Fix critical and high issues (empty descriptions, duplicate names, huge toolsets).
  4. 4Download the markdown report, then generate or refine your MCP server as needed.

What a good vs weak tool looks like

Strong description

list_pets — List pets in the store. Use when the user asks for available pets. Returns a JSON array.

Says when to call it and what comes back.

Weak description

do_stuff — stuff

Vague name + thin description → models guess wrong.

Accepted inputs

MCP tools/list · OpenAI tools · Anthropic tools · OpenAPI 3.x / Swagger 2

All formats scored with the same ergonomics rubric.

Agent tool design tips

  • •Write descriptions for the model: when to use the tool, key args, and return shape.
  • •Describe every property; enums beat free-form strings when possible.
  • •Keep default agent surfaces small—exclude admin/DELETE unless required.
  • •Prefer verb_noun names (list_invoices, create_refund) over vague helpers.
  • •After generating from OpenAPI, re-check here before adding the server in Cursor.

Frequently Asked Questions

What does the MCP Tool Checker evaluate?

It scores agent ergonomics: tool descriptions, parameter schemas, naming quality, destructive/sensitive surface area, and estimated schema token cost. The goal is fewer wrong tool calls from Cursor, Claude, and other MCP clients.

What formats can I paste?

MCP tools/list JSON, a raw tools array with inputSchema, OpenAI function/tools JSON, Anthropic input_schema tools, or a full OpenAPI/Swagger spec (JSON or YAML). Everything is parsed in your browser.

Is this the same as running an MCP server compliance probe?

No. Runtime probes test handshakes and live servers. This checker focuses on static tool-schema quality—the #1 reason agents mis-call tools—without installing a CLI or starting a process.

Does my tool schema leave my browser?

No. Parsing and scoring run 100% client-side. Instant Tools never uploads your MCP manifest or OpenAPI file.

How is this related to OpenAPI to MCP?

Use OpenAPI to MCP to generate a server, then paste the tools (or the original OpenAPI) here to validate agent readiness before wiring Cursor. The two tools share the same browser-local privacy model.

Is the MCP Tool Checker free?

Yes. No signup, no usage limits, and no watermark on the markdown report.

Related Tools

OpenAPI to Tool Schemas

Zod, JSON Schema, OpenAI & Anthropic from OpenAPI

Cursor Rules Generator

Generate .mdc, AGENTS.md, and CLAUDE.md for Cursor

OpenAPI to MCP

Generate a Cursor-ready MCP server from OpenAPI

JSON Formatter

Validate tools/list JSON before checking

Hash Generator

Checksum manifests and release artifacts

Prompt & Rules Linter

Score system prompts and agent rules for clarity