Stability and Bug Fixes
Released on September 8, 2026, n8n 2.39.0 resolves a large set of issues across the platform. Core workflow engine fixes ensure that scheduled runs are marked correctly when workflows are unpublished, task runner timeouts respect shutdown windows, multi-input nodes execute in the expected order, webhook nodes resume all output branches, and detached sub-workflows finish before workers stop. Credential and security-related fixes include redacting OpenTelemetry header values in settings responses, properly checking publish permissions, and redirecting failed OIDC logins. API and editor fixes address permission checks on republish, credit balance alignment, workflow checksum refreshes, and AI Assistant draft restoration. Node-specific fixes cover AMQP Trigger reattachment, Anthropic and Google Gemini model handling, Freshservice and Jira request formatting, LDAP filter escaping, Microsoft SQL TLS connections, Microsoft Teams UTC meeting times, WordPress OAuth2 credentials, and many others.
New Features and Integrations
This release introduces several capabilities for self-hosted users. A new public API endpoint handles workflow versions, and source control now includes push, pull, status, and project management endpoints. Instance AI gains folder exploration, conversation search, credential selection guidance, opt-in concurrency limits, and reporting. The AI Builder becomes capability-aware for agent modifications, and chat triggers can start new agent sessions. Encryption improvements cache keys in memory and make write paths pluggable. MCP registry support now accepts existing credentials, and Databricks gets a hidden chat model sub-node with partner user-agent headers and templated MCP URLs. New node features include Atlassian Service Account authentication for Jira and Confluence, G Suite Admin org unit path options, GraphQL predefined credentials, Microsoft Teams channel message get/reply and Online Meeting operations, plus Edit Image text alignment options.
Performance Improvements
Two performance upgrades are included: source control push now bounds peak memory usage, and the Supabase node fetches its API schema once instead of once per column dropdown, reducing redundant calls and improving UI responsiveness.
