Skip to content
IsoKron

Sibling tools have distinguishable descriptions

When two tools have nearly the same description, the model cannot tell which to pick; say in each description what sets it apart from the other.

Check design-near-duplicate · Schema and description quality · default severity: attention · rubric 2.1.0

What fails

Two tools have nearly the same description (word overlap of 85% or more), so the model cannot tell when to pick which.

The rule

Each tool description should state precisely what the tool does and when to invoke it.

Source: Anthropic connector pre-submission checklist

How to fix it

Say in each description what sets it apart from the other.

  1. Find pairs that differ in one word (list_issues and search_issues described the same way).
  2. State the difference: input (ID vs text), scope (mine vs all), output (summary vs full).
  3. Merge the tools if there is no real difference.

TypeScript SDK

TypeScript
import { McpServer } from '@modelcontextprotocol/server';
import * as z from 'zod';

const server = new McpServer({ name: 'issues', version: '1.0.0' });

server.registerTool(
  'list_my_issues',
  {
    title: 'List my issues',
    description: 'List issues assigned to the signed-in user, newest first. Use for "my issues"; for everyone’s issues use list_issues.',
    inputSchema: z.strictObject({
      state: z.enum(['open', 'closed', 'all']).default('open').describe('Which issues to include'),
    }),
    annotations: { readOnlyHint: true, destructiveHint: false, openWorldHint: false },
  },
  async ({ state }) => ({ content: [{ type: 'text', text: 'Done' }] }),
);

Python SDK

Python
from typing import Annotated, Literal

from pydantic import Field
from mcp.server.mcpserver import MCPServer
from mcp.types import ToolAnnotations

server = MCPServer("issues")

@server.tool(
    title="List my issues",
    annotations=ToolAnnotations(readOnlyHint=True, destructiveHint=False, openWorldHint=False),
)
def list_my_issues(
    state: Annotated[Literal["open", "closed", "all"], Field(description="Which issues to include")] = "open",
) -> str:
    """List issues assigned to the signed-in user, newest first. Use for \"my issues\"; for everyone’s issues use list_issues."""
    return "Done"

Run against @modelcontextprotocol/server 2.4.0, mcp 2.3.0 and fastmcp 4.2.2 on 2026-11-18; the checker confirms this check passes on the tool list each snippet produces.

Questions

How does the checker measure similarity?

Word overlap between descriptions; 85% or more is flagged. Generated tools often share a template and trip it.

Should I merge similar tools?

If they differ only in a filter, one tool with that filter as an argument is usually better for the model.

Further reading

How this check is graded

It counts toward: Schema and description quality. See the rubric.

Last verified against Anthropic connector pre-submission checklist on 2026-11-18. The rule above is quoted in the source’s own words. This page as Markdown.

Check your tool list: paste your tools/list, nothing is sent anywhere