n8n beta

2.39.0 (2026-09-08)

Bug Fixes

• ai-builder: UI tweak to prevent overflowing text on err callout when on Agent Preview (#37615) (c2e8d02)
• AMQP Trigger Node: Reattach the receiver link when the broker detaches it (#37431) (ddf6534)
• Anthropic Chat Model Node, AI Agent Node: Send disabled thinking to the provider and parse content-block output (#37954) (309fa8d)
• API: Check publish permission when a workflow save republishes (#37264) (485e370)
• Chat Trigger Node: Fix singular copy on the connect bar (#37945) (369c9e2)
• Chat Trigger Node: Re-authenticate hosted chat on session change (#37868) (64ba49d)
• Check constructor-assigned and type-asserted node descriptions (#37523) (e23ee05)
• Code Node: Keep identity and reporting fields when wrapping task runner errors (#37490) (b0550cb)
• core: Allow HashiCorp Vault external secrets to read a KV v2 sub-path (#37748) (fed5d70)
• core: Allow webhook nodes to set Set-Cookie response headers (#37883) (6828320)
• core: Always redact OTel header values in settings responses (#37283) (4b0de25)
• core: Cap task runner task timeouts to the graceful shutdown window (#37651) (9a64820)
• core: Complete a scheduled occurrence as not dispatched when its workflow is unpublished (#37721) (29c7d9f)
• core: Ensure running job cleanup when a workflow run rejects (#37621) (51ac147)
• core: Execute multi-input nodes re-fired through a loop-back edge in legacy order (#37561) (903fe57)
• core: Fix Azure OpenAI classic and Foundry credential setup and execution for agents (#37251) (731ea42)
• core: Forward all output branches when a waiting webhook node resumes (#31714) (7c21f4b)
• core: Give every simulated node at least one plausible item (#37557) (918067a)
• core: Harden agent workflow tools against unpublished, missing, or incompatible workflows (#37772) (1994332)
• core: Hide instance MCP OAuth discovery when MCP access is disabled (#36726) (9768c1b)
• core: Keep a serving external secrets provider active when its replacement fails (#37781) (217c7b3)
• core: Keep Instance AI data-table queries targeted while diagnosing (#37489) (df4d30b)
• core: Keep reporting the original job error when the job key is gone (#37993) (7ce17be)
• core: List the valid techniques when get_workflow_best_practices rejects input (#37580) (53b1ad7)
• core: Make pending agent creation and thread binding idempotent (#37074) (085a7a9)
• core: Make scoped workspace path errors actionable for agents (#37887) (03e1793)
• core: Mark retired node types as deprecated in their workflow-sdk type definitions (#37628) (cf0ef47)
• core: Prefer service-specific credentials for HTTP requests (#37844) (500ea39)
• core: Prevent Anthropic agent threads from breaking permanently and keep run errors visible (#37640) (dc5f420)
• core: Redirect failed OIDC logins to the sign-in page (#37428) (3e09299)
• core: Report the provider response when a chat model returns no choices (#37959) (49e7821)
• core: Return 503 when the AI assistant service is unreachable (#37957) (36426a8)
• core: Stop npm audit from stalling Instance AI sandbox setup (#37864) (7ac335d)
• core: Stop task broker before task runner processes on shutdown (#37833) (2ad8d01)
• core: Stop the AI Assistant from pulling the whole workflow through the model for small edits (#37644) (8ef5ba6)
• core: Take archived workflows into account during package import (#37668) (98dfb30)
• core: Update Qdrant client and remove undici v6 dependencies so AI requests work on Node 26 (#37758) (325feb3)
• core: Wait for detached sub-workflows before the worker stops its task runner (#37737) (afc4a54)
• core: Workflow tools should wait and not throw if parent WF is suspended (#37087) (10ea0c8)
• editor: Align credit balance copy to "left" across views (#37423) (9c131d6)
• editor: Expand AI question text input automatically (#37627) (5e3774a)
• editor: Hide Gateway credits for HTTP Request and GraphQL nodes (#37810) (c292476)
• editor: Preserve n8n credits when copying workflows (#37253) (8946168)
• editor: Refresh workflow checksum after publishing from an embedded editor (#37949) (8119ea3)
• editor: Restore Instance AI draft when thread creation fails (#37874) (0820083)
• editor: Scope end-user credential trigger warning to reachable nodes (#37081) (8399317)
• editor: Show restart notice after license activation (#37841) (f56d87d)
• editor: Show skill size limits as a character count (#37847) (a9d5fb7)
• editor: Update Agent builder assistant placeholder (#37529) (20cf19f)
• Fix attribution footer link on the Form Ending page (#37288) (e08a4fd)
• Freshservice Node: Send change filters as dedicated query params (#34380) (5ae1861)
• Google Gemini Node: Validate domain before downloading (#37459) (a1c1f6f)
• Guarantee package.json restore after the FE trim step (#37446) (fd6db75)
• Jira Node: Strip trailing slashes from domain URL (#32163) (a69155a)
• Jira Software Node: Send expand as comma-separated string for Jira Cloud searches (#37527) (9f2ebe5)
• Ldap Node: Escape expression output in the Custom Filter field (#37301) (91f8794)
• Microsoft SQL Node: Fix TLS error when the IP is used to connect to the server (#37870) (5ddfb60)
• Microsoft Teams Node: Send Online Meeting times in UTC and share the Graph helpers (#37723) (d5e05b2)
• Mistral Cloud Chat Model Node: Handle Request objects in undici-backed proxyFetch (#37946) (7533aaa)
• Read/Write Files from Disk Node: Allow glob character classes in File(s) Selector (#36901) (3c48b53)
• Sign Microsoft Dynamics OAuth2 requests with the access token (#37855) (6fa5b3d)
• Update MCP server status in AI Assistant when tool call failed or credentials updated (#37197) (b639fe5)
• Use consistent personal space icon in the Instance AI artifacts (#38011) (cbaa0be)