N8n 2.40.0

2.40.0 (2026-09-15)

Bug Fixes

• AI Agent Node: Preserve Anthropic thinking blocks with empty text across tool calls (#38302) (7ac975c)
• ai-builder: Enhance guidance UX for agents with no sessions on builder session tab (#38243) (1ed3a24)
• ai-builder: Guard against 429 errors on channel integrations (#38106) (93ca2d1)
• ai-builder: Show reusable credentials in the Instance AI agent credential card (#38276) (999b541)
• Allow one icon file for the light and dark icon slots (#37711) (9095e9a)
• API: Let global members grant folder scopes to API keys (#37654) (aa91e71)
• API: Return null variable values as empty string & fix newly imported source control variable values (#38488) (b9a7f03)
• Azure OpenAI Chat Model Node: Resolve the proxy against the Azure endpoint so NO_PROXY applies (#38026) (7383fd7)
• Bump task-runner-launcher to 1.5.0 (#38309) (9135888)
• Center the registration badge in the Usage and Plan heading (#38275) (cf6cb77)
• Close the variable modal on a backdrop click (#38286) (45a7f74)
• Confluence Node: Match hyphenated and multi-word titles in the page picker search (#38175) (fdb31d2)
• Confluence Node: Return the page sub-tree in tree order and skip drafts (#38403) (b0f47eb)
• core: Accept null headers from mcp registry api (#38614) (02ee221)
• core: Accept numeric pagination tokens from resource locator list methods (#38147) (7ab1582)
• core: Add model boundaries to external tool results (#37316) (d328bbf)
• core: Allow a fallback model alongside the primary on AI nodes (#38007) (2550501)
• core: Apply the SSO expression mapping env var in the provisioning loader (#37170) (b4fee56)
• core: Ask which MCP connector to set up before the AI Assistant gives a URL (#37992) (899d52d)
• core: Batch LangSmith agent session exports (#38030) (79d936f)
• core: Bound the wait for the workflow publication lock (#38019) (5212777)
• core: Check Gateway credits eligibility against the correct sub-node (#38193) (9823c2e)
• core: Consolidate secret redaction in @n8n/utils and redact more credential shapes (#38054) (c5653f6)
• core: Correct the expression prefix warnings on fields that cannot hold an expression (#37791) (c7c179b)
• core: Count executions that crash recovery marks as crashed (#37733) (3521e54)
• core: Do not persist an execution when preExecute throws (#37701) (53ecf6b)
• core: Enforce execution timeout for stuck jobs in queue mode (#36258) (92f702d)
• core: Execute MCP toolkit members on workers (#38069) (c0f6471)
• core: Forward the receiver in the lazy-proxy set trap (#37321) (b3306b2)
• core: Ground provider OAuth scope guidance in n8n docs (#38066) (2afb5b6)
• core: Keep a resumed execution in the trace of the parked execution (#38033) (5414403)
• core: Keep node settings, positions and error routes through an assistant workflow edit (#38254) (c95756d)
• core: Keep nodes clear of the agent card in SDK-generated workflow layouts (#38093) (4b4d670)
• core: Memoize pairedItem ancestry traversal (#38085) (f3db19d)
• core: Merge system messages for strict providers (#37908) (19651de)
• core: Prefer native nodes over Code nodes in n8n Assistant workflow builds (#38598) (4e1fcf6)
• core: Preserve link URLs in inbound agent chat messages (#38294) (2beccad)
• core: Prevent false agent model stream stalls and show run errors in chat history (#37558) (26afe95)
• core: Re-register the n8n Trigger on every publish through the publication service (#38153) (cbda01a)
• core: Read the Instance AI folder-exploration flag as a variant, not a boolean (#38479) (4579e78)
• core: Recognize always-loaded tools in agent tool discovery (#38204) (4a6e625)
• core: Reconnect the Email Trigger (IMAP) on every publish through the publication service (#38190) (db60f99)
• core: Reconnect the Postgres Trigger on every publish through the publication service (#38211) (0c3c9d7)
• core: Redact execution data on partial private-credential runs (#38364) (840aef6)
• core: Redact opaque field values in confirmed secret containers (#38067) (3bd0e69)
• core: Register redaction context hook on all instance types (#37298) (56383fc)
• core: Repair legacy-format data-encryption keys during bootstrap (#38100) (4b36c51)
• core: Require verification coverage for every workflow trigger (#37961) (429116f)
• core: Retry Instance AI model turns when the provider stream closes without output (#38080) (b6b5ecd)
• core: Return Instance AI node output grouped per output branch (#38414) (e44fbd3)
• core: Return luxon values from expressions as instances (#38034) (0c6c9e3)
• core: Reuse an existing consent grant after the visitor authenticates mid-flow (#38084) (3b6faa2)
• core: Route "trigger my workflow" to the existing workflow, not the builder (#38073) (219402c)
• core: Run tool nodes of an AI Agent upstream of the destination node (#38348) (40b6d84)
• core: Scope data table Overview access by project role, not a global gate (#37676) (a47f19b)
• core: Scope end-user credential checks to reachable nodes (#36970) (37e50b5)
• core: Select sandbox npm install flags by provider (#38358) (ab101d5)
• core: Send a JSON Accept header in the Brave Search credential test (#38594) (2f55165)
• core: Settle answered confirmation cards in the AI Assistant run-sync frame (#38328) (4890d04)
• core: Stop MCP workflow updates from dropping unknown operation fields (#38148) (a1f8405)
• core: Stop the workflow SDK collapsing inline branch chains to their branching node (#37717) (37efe11)
• core: Treat empty strings as empty in Data Table filters (#32887) (0c82743)
• core: Treat Instance AI model stream stalls as dropped provider connections (#38079) (b7a58c2)
• core: Trust the CA from GIT_SSL_CAINFO for source control HTTPS remotes (#38258) (e803d4b)