agentgateway 1.6.0 in production: 0.2-0.6ms latency, <35MB idle, rate limits hidden in HTTP 200
ghumare64 · x · 2026-10-14
Developer Rohit Ghumare ran a production-grade test of agentgateway 1.6.0, the Rust proxy donated to the Linux Foundation that fronts MCP servers, A2A agents, and LLM APIs from one endpoint.
- Key numbers: added latency 0.2-0.6ms, idle memory under 35MB, and rate limit errors wrapped in HTTP 200 instead of 429 — clients never see them unless they check
- The gateway also hides a denied tool call behind the same error as a missing one, making "no access" vs "doesn't exist" hard to distinguish
- One endpoint replaces per-agent lists of MCP server addresses and credentials, centralizing auth, tool permissions, and call auditing
- Deep dive into the 2026-07-28 MCP protocol change: version now declared per request (meta + MCP-Protocol-Version header), enabling stateless forwarding by any gateway replica
- The 25-min guide builds a setup step by step, measuring each policy's cost on real requests
More from coding & agent
- Codex autocompaction gets stuck: one steer and 10 million tokens later it's still parroting stale state — Aryvyo · 2026-10-14
- Agentic engineering vs AI engineering: Hugo Bowne-Anderson's clean split — Al_Grigor · 2026-10-14
- PostHog to devtools: stop asking 'how to add AI,' ask 'what's the agent-first version' — edwin · 2026-10-14
- Devin SWE-2 jumps to #3 on attention-to-detail coding test, tying Claude Sonnet 5.5 — silasalberti · 2026-10-14
- MCP hits ~500M monthly downloads but stays free — the money goes to key-management security firms — FinanceYF5 · 2026-10-14
- One-shot flying dragon: Gemini 4 Argon with Blender MCP and Tripo 3D builds a pure .blend scene — sidahuj · 2026-10-14