rpgeeganage/lazy-tool

local-first MCP discovery runtime for agents — search before invoke, reduce prompt bloat, and route to local MCP tools

28
/ 100
Experimental

As a developer working with AI agents and Model Context Protocol (MCP) servers, you might find that connecting many tools leads to slow responses and increased token costs because the agent reads every tool's schema. This tool acts as a local proxy, intelligently routing agent requests to the correct MCP tool without sending every schema to the model. It takes your existing MCP server configurations (from IDEs or custom YAML) and provides a single endpoint for your AI agent, outputting faster, cheaper, and more accurate tool use.

Use this if you are a developer building AI agents that interact with multiple MCP tools and want to reduce prompt bloat, improve agent accuracy, and lower token costs without managing complex infrastructure.

Not ideal if you only use Claude Code's built-in tool search and do not work with other providers or local MCP servers.

AI-agent-development Model-Context-Protocol developer-tooling AI-workflow-optimization local-deployment
No Package No Dependents
Maintenance 13 / 25
Adoption 6 / 25
Maturity 9 / 25
Community 0 / 25

How are scores calculated?

Stars

19

Forks

Language

Go

License

MIT

Category

discovery

Last pushed

Mar 31, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/mcp/rpgeeganage/lazy-tool"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.