Source: github.com/n8n-io/n8n/releases/tag/n8n%402.46.0
The release adds Microsoft Foundry deployments to the Agents model selector and lets the Agent Builder set up Foundry models. Claude deployments on Microsoft Foundry are also supported in Agents, and Foundry deployments are priced by the model they serve.
Several fixes target agent behavior. Bedrock auth is now pinned and the AWS region is validated in agents, and reasoning is replayed to custom OpenAI-compatible models. SDK memory threads are deleted when an agent is deleted, and custom OAuth2 scopes are now allowed.
Other changes cover MCP and tracing. Webhook and MCP Trigger nodes are authenticated through trusted sources, the W3C trace context is sent on MCP server requests, and OpenTelemetry GenAI attributes are added to AI Agent spans. The changelog also lists mention-only Slack replies for agents.
Why it matters if you build agents
Teams users and Foundry-hosted Claude and other models can now be used directly in n8n agents, which removes some custom wiring.
Developers who rely on SDK memory should check their cleanup flow, since memory threads are now removed when their agent is deleted.
Because this is a pre-release, workflows that depend on the new behavior should be tested before they reach production.
Get stories like this every weekday
One short email with the day's agent tooling news.