Onemcp FAQs

Question 1

What is 'Progressive Tool Discovery' and why is it important?

Accepted Answer

Progressive Tool Discovery allows LLMs to efficiently find and load tool schemas only when they are actually needed. Instead of consuming tokens by loading all schemas upfront, LLMs can use 'tool_search' to discover relevant tools and then request detailed schemas as required, optimizing token usage and speed.

Question 2

How does Onemcp achieve 99% token efficiency for LLMs?

Accepted Answer

Onemcp implements a meta-tool architecture that exposes only two generic tools ('tool_search' and 'tool_execute') to the LLM, rather than hundreds of individual tool schemas. Combined with progressive tool discovery, this significantly reduces token overhead, preserving the LLM's context window and lowering operational costs.

Question 3

What is Onemcp and its primary purpose?

Accepted Answer

Onemcp is a universal Model Context Protocol (MCP) aggregator designed to unify multiple external MCP servers and custom internal tools into a single, efficient interface. Its primary purpose is to drastically reduce token usage for Large Language Models (LLMs) when interacting with numerous tools.

Question 4

Is Onemcp compatible with different LLMs and MCP servers?

Accepted Answer

Yes, Onemcp offers universal compatibility. It works with any MCP-compliant LLM, including Claude, OpenAI, and Gemini, as well as any external MCP-compliant server. It also supports integrating custom Go-based internal tools.

Question 5

What performance optimizations are included in Onemcp?

Accepted Answer

Onemcp includes several optimizations such as fuzzy search for robust queries, schema caching for external tools, configurable search result limits, lazy loading of schemas at different detail levels, and a hybrid schema approach that provides a complete tool list via a file while returning limited tools inline for discovery.

Onemcp

概要

主な機能

ユースケース