n8n pushed out a cluster of releases across its 2.33 and 2.34 branches in the first week of August, bundling together an Agent Builder capability with materially deeper support for the Model Context Protocol. The headline changes include enabling the 'instance-ai' module by default (removing a startup warning that previously required manual configuration), exposing agent test execution through MCP so external tools can trigger and inspect a workflow's AI agent runs, letting the assistant search n8n's own MCP registry as a callable tool, extending the 'n8n Connect to Agent Builder' node so more building blocks are available when constructing agents visually, and adding support for the newer MCP 2026-07-28 discovery handshake at the instance server level. Practically, this continues n8n's push to be the connective tissue between low-code workflow automation and the growing ecosystem of MCP-compatible AI agents and tools, rather than treating AI as a bolted-on node type. For teams building internal automations, the discovery handshake and registry search matter because they reduce the manual wiring needed to expose or consume MCP servers from inside a workflow, and the default-on AI module lowers the barrier for less technical users to start experimenting with agentic steps. It's an incremental release cadence rather than one big version bump, but the direction — MCP becoming a first-class citizen inside a mainstream workflow tool — is the notable part for anyone standardizing on MCP as their agent-integration protocol.