mcp-top: Track the Real Cost of Your MCPs
ziwi_wiwi · reddit · 2026-07-19
mcp-top is a local analytics tool for agent platforms like Claude Code / Codex / Cursor, designed to identify:
- Which MCP servers are configured;
- How many tools they define and how much context they consume;
- Whether these tools were actually invoked in past sessions.
Its core value is cross-referencing two types of data:
- MCP configurations: Available servers and tools;
- Session transcripts: Actually called tools.
This allows you to rank MCP servers by "context cost vs. actual usage," helping you remove services that consume tokens but are never called.
The post also highlights key design principles:
- Read-only by default; it doesn't modify configs or upload data;
- Uses interval estimation for Claude Code's Tool Search, avoiding the illusion of a fixed token tax;
- Clear visibility into coverage and skip reasons;
- Built purely with Python standard libraries, MIT-licensed, and includes 190 tests.
This is a classic infrastructure and efficiency tool for coding agents.
More from coding & agent
- Agent Field Kit launches $29 security runbooks for production AI agents — alvelda · 2026-10-27
- Dev mixes Claude specs with abliterated GLM to build SS7 threat detector — l4rz · 2026-10-27
- Cloudflare Durable Objects validated as top architecture for AI agent WebSocket servers — tristanbob · 2026-10-27
- GitHub Next's Chopin prototype explores real-time multiplayer agentic planning — max__drake · 2026-10-27
- Veynod: An Open-Source Shared Workspace Where AI Agents Collaborate Like a Team — TKOWNS · 2026-10-27
- TMLR paper: enforced thinking doesn't reliably make LLM agents disclose key info to users — SharonYixuanLi · 2026-10-27