Ai

n8n 1.123.49 Fixes MCP Connection Leak in Webhook Drain

n8n 1.123.49 Fixes MCP Connection Leak in Webhook Drain

n8n dropped a quiet but critical update this week. Version 1.123.49 might look like just another patch bump, but it fixes a bug that could silently degrade AI-powered workflows. The issue? MCP connection leaks during webhook shutdown. If you're using n8n to orchestrate AI agents or model interactions, this one's for you.

What Changed

The fix targets the core webhook close functions. Specifically, the team added proper draining of close handlers to prevent MCP (Model Context Protocol) connections from lingering. The commit (69b2fbd) is small but surgical.

Before this patch, when a webhook stopped receiving requests, its close function could leave MCP connections open. Over time, that meant resource leaks. Think stale sockets, orphaned sessions, and eventually degraded performance.

Why It Matters

MCP is becoming the standard for hooking up AI models to tools like n8n. It's how an agent sends a query to a language model and gets a structured response back. If those connections leak, you're not just wasting memory – you're risking dropped AI responses and unpredictable behavior.

This is especially painful in production. Imagine a customer-facing chatbot that gradually slows down because its n8n workflows are piling up half-open connections. Not a good look. The fix ensures that when a webhook stops, everything behind it cleans up properly.

It's a small patch, but it shows n8n's maturity. They're paying attention to the plumbing, not just the flashy features. For anyone running AI agents at scale, this update is a no-brainer.

Official Source: https://github.com/n8n-io/n8n/releases/tag/n8n%401.123.49