Benchmarking 8 public remote MCP servers: latency, schema token bloat, silent 500s
FreeMembership5107 · reddit · 2026-09-06
The creator of MCP Sentinel ran synthetic protocol tests against 8 public remote MCP servers, measuring handshake latency, JSON-RPC 2.0 conformance, and tool schema token cost. Key findings: 2ools App MCP exposes 81 tools with a 133.7 KB schema (33,450 tokens injected into the system prompt before you type anything), vs 1,980 for AWS Knowledge; handshake latency ranged from 283ms (AWS) to 2.8s (HumanForAI); Memeboat fails initialize with an HTTP 500 that fails silently in Cursor and Claude Desktop; some servers return 200 OK with empty tool arrays. Tooling is open source, and the author asks whether people prune tools dynamically or let clients eat the context overhead.
More from coding & agent
- Codex over SSH fails on headless Windows: sandbox runner can't start in Session 0 — Chance-Struggle-6285 · 2026-09-06
- Devs debate multi-machine agent networks: trust and policy, not messaging, is the hard part — No_Praline7219 · 2026-09-06
- Should you post-train a model for your agent harness? Dev seeks real cost data — Small_Luck8177 · 2026-09-06
- LLMPvP: an MCP arena where LLM agents duel in ranked Chess and Go with Glicko-2 — vudueprajacu · 2026-09-06
- Astra vs Fable 5.1 on real ML tasks: Astra grinds agentically, Fable writes better — returnity · 2026-09-06
- Where coding agents beat human-in-the-loop: inventing strategies for brand-new HPC algorithms — IgorCarron · 2026-09-06