github n8n-io/n8n n8n@2.45.0

latest releases: stable, beta
pre-release3 hours ago

2.45.0 (2026-10-20)

Bug Fixes

  • AI Agent Node: Replay Anthropic thinking turns with their tool arguments (#41241) (ea98bd0)
  • Align data table and variable editing on protected instances (#41258) (d4bde14)
  • Allow skipping table initialization (#40437) (d245b17)
  • API: Apply projectId filter to executions for member API keys (#40887) (3075b60)
  • API: Log errors before returning generic 500 in Public API (#41186) (fe36279)
  • Check provisioning license on Users settings page (#41048) (066480d)
  • Convert to File Node: Decode data URIs and validate Base64 input (#38554) (315e0a5)
  • core: Allow gateway model IDs and forward credential headers in agents (#40470) (791c7bc)
  • core: Allow project promotion without an instance export (#40939) (e0fe040)
  • core: Allow switching a node back to a credential the user cannot use after a save (#41198) (e7f0607)
  • core: Clear agent runtime cache when credential is updated (#40961) (4cb7a6e)
  • core: Close agent after run when invoked from workflow (#41128) (f596162)
  • core: Confirm project deletions before instance apply (#40941) (f631112)
  • core: Create an XML parser for each webhook request body (#41098) (1950434)
  • core: Describe MCP servers only from user or registry text in Agent Builder (#40974) (248c863)
  • core: Display an error message when failing to connect to oauth credentials (#41058) (952b1e7)
  • core: Evict idle mcp-browser sessions to prevent connection leak (#36850) (efd4e21)
  • core: Handle every escaped closing bracket in an expression (#40487) (b09efb7)
  • core: Improve base URL check for declarative nodes (#40778) (e45a65a)
  • core: Keep cancelled Assistant activity settled (#40786) (d74f3f2)
  • core: Keep Observer detail lines instead of discarding the response (#41134) (48edbef)
  • core: Keep offline_access when a protected resource omits it (#41384) (9675910)
  • core: Keep pending setup for nodes that n8n Assistant added in an earlier build (#41070) (50c4ced)
  • core: Keep the workflow source file when n8n Assistant sends blank source (#41201) (b78913f)
  • core: Key each OAuth authorization request to its own session (#39930) (3c447b4)
  • core: Key queue job outcomes by job on shutdown resume (#40569) (6e04287)
  • core: Let LLM failed-attempt handling retry and honour Retry-After (#41002) (67c1981)
  • core: Let preview chat agents make changes through their tools (#41403) (8323dfd)
  • core: Log custom data truncation only above the 512-character limit (#40899) (e74f1c1)
  • core: Pass reasoning effort on open router model node (#41179) (683fd11)
  • core: Redact a parent execution if any sub-workflows have redaction enabled (#41129) (c70b5d4)
  • core: Redact the parent execution when a sub-workflow uses a credential the viewer cannot use (#41085) (a57c8f3)
  • core: Reflect failed publications in the active workflows endpoints (#40641) (96eff1e)
  • core: Remove unused N8N_SCHEDULER_MIN_INTERVAL setting (#41194) (c839d6c)
  • core: Reuse connections in the log streaming webhook destination (#41261) (1581524)
  • core: Skip resource mapper validation when the mapped value is null (#41069) (898701b)
  • core: Stop asking to set up a node when n8n Assistant only adds a step next to it (#41067) (578eea4)
  • core: Trace verification helpers and provider attempts (#40936) (f73dc33)
  • core: Turn off expression spans by default (#41211) (842ec5b)
  • core: Use the module registry to decide if the Assistant has the Agents features (#41026) (726a582)
  • core: Web search misconfigs in agents as UserError (#41257) (f844135)
  • Databricks Node: Bring the AI Agent panel to parity with AI Agent 2.2 (#40989) (e859aba)
  • Databricks Node: Rename Genie resource labels to Genie Agent and the MCP tile to Databricks Genie One (#40760) (14c7790)
  • editor: Apply the standard settings layout to the Context preferences page (#40716) (b3e8634)
  • editor: Cancel the Assistant thread when the hand-off navigation fails (#41056) (d26f914)
  • editor: Cleanup legacy button variants (#41175) (ba97e46)
  • editor: Display custom evaluation metrics correctly (#40852) (f6b24d0)
  • editor: Improve destination validation and surface save errors (#40897) (eaf47cf)
  • editor: Improve NDV credential options performance on large instances (#40198) (b98640f)
  • editor: Keep pending navigation when workflows list query is unchanged (#41125) (116a81e)
  • editor: Keep the used-in rows on the credential sharing tab when personal-space sharing is disabled (#41207) (ca9661c)
  • editor: Keep workflow name in edit mode after Rename menu action (#41120) (fc9f957)
  • editor: Load credential types on startup (#41400) (9a025a1)
  • editor: Match the credential dropdown width to its field and remove horizontal scroll (#41260) (51e53ad)
  • editor: Offer the managed OAuth choice when the app is supplied (#40284) (14d2d30)
  • editor: Place sticky notes and nodes at the right-click and clear the selection (#41365) (28602ed)
  • editor: Prevent doubled base path when opening folders with N8N_PATH set (#33065) (a32c652)
  • editor: Rename Admin Panel to Cloud console (#41187) (8120328)
  • editor: Restore From AI button corner radii (#41176) (20ea671)
  • editor: Restore the new session shortcut in agent chat (#40907) (ebf078b)
  • editor: Return to the requested page after SSO sign-in (#41151) (4a5fdd5)
  • editor: Save only the OpenTelemetry status when enabling or disabling (#41142) (1903d53)
  • editor: Show follow-up run events in the Assistant after a reconnect (#40883) (ffda639)
  • editor: Start the empty-state icon cards cycling on load and use them on the MCP and external secrets pages (#40232) (1f223c8)
  • editor: Stop auto-selecting the Evaluation Trigger over other triggers (#38883) (8b5ea6d)
  • editor: Stop credential icon overlapping the credential name in the agent credential picker (#41262) (70f0131)
  • editor: Unify documentation links across settings pages (#40289) (76cce99)
  • editor: Use the deployment name on the Az Foundry path (#41103) (b3e8b7d)
  • Email Trigger (IMAP) Node: Check the mailbox when the trigger activates (#41013) (a6719b8)
  • Email Trigger (IMAP) Node: Find new mail on servers with a broken IMAP4rev2 mode (#41018) (236220b)
  • ERPNext Node: Return each row once and tolerate empty Get Many options (#36860) (3968a60)
  • Execute Command Node: Don't spawn console on Windows (#40909) (52aed7a)
  • GitlabTrigger: use NodeApiError.httpCode and safe description access in webhook checkExists (#35486) (054288e)
  • Gmail Node: Keep the References header on a draft created in a thread (#41392) (06ba85c)
  • HomeAssistant Node: Only sent Home Assistant attributes on upsert if set (#36298) (b8780d8)
  • Keep free-floating sticky notes clear of workflow nodes (#41195) (131e9c8)
  • MCP Trigger Node: Pass through MCP content arrays instead of double-wrapping tool results (#39131) (f483ebd)
  • Microsoft Teams Node: Add the TeamMember.Read.All scope and correct the Trigger scope notice (#39819) (ddc9769)
  • OpenAI Node: Report token usage for image operations (#38463) (a3f4711)
  • OpenAI Node: Send max_completion_tokens when it analyzes an image (#38656) (e972309)
  • Pass structured inputs to OpenAI Responses tools (#40638) (1d2c879)
  • Postgres Node: Keep comma-containing values as one query parameter (#40240) (f5dc06f)
  • Restore command bar focus and standardize creation icons (#41185) (9e3b52c)
  • Send Email Node: Send to every recipient when a recipient field is an array (#36467) (660849d)
  • Skip dependency files and continue after custom node load (#41117) (5a3fc39)
  • Splunk Node: Preserve errors when continuing after failed items (#40368) (e29c3a4)
  • Validate attendee values (#40953) (21221ad)

Features

  • API: Honor Idempotency-Key on Public API POST requests (#40978) (b7b3d72)
  • API: Return the parent folder on workflow read endpoints (#40981) (a906d2f)
  • Azure OpenAI Chat Model Node: Add an optional Project field to the Foundry credentials (#41181) (0c1a787)
  • core: Add an endpoint that notifies workflow owners about migration findings (#40952) (d8a0666)
  • core: Add an optional model for n8n Assistant memory tasks (#41137) (f34c093)
  • core: Add create-with-values credential missing mode to package import (#36767) (d3bc275)
  • core: Add self-healing results to Inbox (#40741) (1405d4a)
  • core: Add the blocking project and opt-in state to policy violations (#40985) (4419fc4)
  • core: Add typing indicator, media download and media retention … (#40920) (adedbbf)
  • core: Add WhatsApp credential check and webhook handshake status (#40921) (9a524e7)
  • core: Append node tool attributions to agent replies and add Genie attribution and deep links to the Databricks node (#40765) (44c8a31)
  • core: Apply budget alerts and stops to sub-agent run (#40884) (9498088)
  • core: Enable node and credential type policies by default (#41022) (dd79a61)
  • core: Enable the durable scheduler by default (#40891) (f6861fe)
  • core: Implement automated migrations for removed nodes (#40877) (739c59e)
  • core: Only show execution input and output for agent subworkflows in timeline (#41189) (4d292a6)
  • core: Refuse revoked access tokens of the local server on the inbound pipeline (#41266) (0936dfd)
  • core: Remove the startup pass from workflow history compaction (#40886) (45eea1b)
  • core: Report the CPU and memory limits n8n was given in telemetry (#40421) (d6d24f2)
  • core: Report the runtime n8n runs on in telemetry (#40420) (b2edcb5)
  • core: Run the Form and Chat trigger redirect login against a trusted source (#41123) (862c7ec)
  • core: Store promotion exports in n8n directory (#40863) (333a010)
  • core: Surface the responding model name in LLM sub-node run data (#41161) (b10c97e)
  • core: Suspend in-flight production executions at worker shutdown (#40568) (f3e2bd8)
  • core: Treat notified migration findings as unresolved (#40951) (347385b)
  • Databricks Node: Show the Vector Store, Embeddings and AI Agent nodes in the node picker (#41206) (191ea98)
  • Don't save MCP connections without user approval (#39996) (d27de60)
  • editor: Align the WhatsApp channel setup and settings with the … (#41283) (275f7eb)
  • editor: Extend N8nCopyInput with accessibility, redaction, loading and actions (#34942) (c084d99)
  • editor: Let admins notify workflow owners from the migration rule detail (#40949) (a972eeb)
  • editor: Let admins notify workflow owners in bulk from the migration rule detail (#41060) (de42c95)
  • editor: Link to the credential sharing tab when execution data is redacted (#40976) (da3491f)
  • editor: Publish agent modal and modal for warning about unpublished workflows (#40987) (e47940e)
  • editor: Rework the WhatsApp channel setup to match the new design (#40923) (af0202d)
  • editor: Show agent personalisation icon in subagent modal (#41094) (6160624)
  • editor: Show multiple workflows in the review detail view (#40695) (f6d70c1)
  • editor: Show policy-blocked items in the source control pull modal (#40604) (1ea18e9)
  • editor: Show the model that answered in the AI log header (#40988) (caeaff7)
  • editor: Show the policy reason when an agent tool call is refused (#40673) (bb45bca)
  • editor: Update budget panel when agent executes (#40971) (632d8b9)
  • editor: Update UI with n8n rebrand (#40342) (bffe7d8)
  • Implements workflow previews above Assistant empty state experiment (#40913) (eac2f46)
  • Kafka Node: Make version 2 the default for new Kafka nodes and triggers (#40744) (e5866ac)
  • Monday.com Node: Add V2 with full V1 parity as the default (#35515) (8977833)
  • Overhaul MCP tool permissions for Agents (#39878) (bda9118)
  • Show Gemini models by name and list prioritized models first in model pickers (#40965) (acf5291)
  • Text Classifier Node: Add opt-in confidence scores to the output (#40743) (3800bce)

Performance Improvements

  • editor: Load prettier only when the user formats code (#41014) (e2529a8)

View guided diff Turn on auto-fix

Don't miss a new n8n release

NewReleases is sending notifications on new releases.