github n8n-io/n8n n8n@2.36.0

latest releases: stable, beta
pre-release4 hours ago

2.36.0 (2026-08-18)

Bug Fixes

  • ai-builder: Align HITL approval resume schemas with confirm envelope (#36047) (be1fde1)
  • ai-builder: Enable Save in agent tool modal when credential auth is selected (#36012) (4fe94d8)
  • ai-builder: Prevent agent from claiming integration tool call success early (#36196) (547cfb9)
  • ai-builder: Recover stuck tool-call UI and MCP timeout progressive update (#36162) (e1cd757)
  • ai-builder: Surface existing LLM-provider credentials when a filtered lookup finds none (#35833) (d73d625)
  • ai-builder: Surface skill save validation errors and stop reverting skill renames (#36055) (b08cde6)
  • Avoid false port-in-use error in get-n8n script on Windows/WSL (#36026) (6f2033f)
  • core: Accept Always allow scope on data-tables resume (#36070) (560455f)
  • core: Allow details field in continueErrorOutput mode (#35939) (55927c4)
  • core: Apply TLS options per hop when requests go through a proxy (#35518) (68bcdfe)
  • core: Bound database pool teardown so connection recovery always completes (#36013) (da314d3)
  • core: Bound workflow publication outbox record processing with an abort deadline (#36197) (c0278b0)
  • core: Crash enqueued executions with unreadable data (#36010) (fd67bf8)
  • core: Detect node settings changes when comparing workflow versions (#36066) (e4f0c33)
  • core: Execute tools in the order the root requested them (#36009) (acd4fda)
  • core: Explain Browser Use failures caused by other browser extensions (#36011) (f66a0e8)
  • core: Expose incoming request headers to MCP trigger tools (#35236) (64ed864)
  • core: Fix non existent workflow id pathway for AIA (#35984) (6ab2569)
  • core: Fix z-indexing of expression modal over agent NDV (#36233) (1fa6ef0)
  • core: Gate AI Assistant workflow publish offers on live verification (#36394) (602e6a0)
  • core: Handle dropped AI provider connections as transient (#35587) (72235d3)
  • core: Handle oversized chat attachments gracefully (#36061) (de8b358)
  • core: Harden agent HTTP Request tool configuration (#36463) (0791359)
  • core: Improve agent builder skill and channel handling (#35980) (a723e79)
  • core: Keep node IDs stable when the AI Assistant edits a workflow (#36236) (a71e45a)
  • core: Keep resolving end-user credentials after a trigger stops listening (#35973) (a447ac4)
  • core: Keep the account panel on hosted forms once every account is connected (#36429) (a8754b4)
  • core: Let Instance AI run a specific trigger's flow (#36410) (1a7425d)
  • core: Make published agent channel add, replace, and remove failure-safe (#36100) (79c1393)
  • core: Prefer managed OAuth auth types when the AI builder resolves credentials (#35929) (1b4c0ba)
  • core: Preserve Anthropic thinking history across agent turns (#36185) (5f6d734)
  • core: Rebuild the node registry without emptying it first (#36387) (86facbd)
  • core: Recover model turns that end without usable output (#36423) (bae82bd)
  • core: Remember skipped Instance AI credential setup for the rest of a thread (#36308) (6ace2a5)
  • core: Report the real activation mode for triggers activated via the publication outbox (#36123) (1a12af6)
  • core: Report the workflow's parent folder in get_workflow_details (#36234) (9b97fb5)
  • core: Require trigger node name for MCP workflow executions (#36341) (7ea9d54)
  • core: Resolve agent credentials through the node decryption path so external secrets work (#36275) (9cbbd59)
  • core: Resolve AI Assistant sticky note placement after layout (#36320) (b671cf4)
  • core: Resolve Custom API Call support through inherited credential types (#35820) (88efdb9)
  • core: Resolve operation defaults per resource in generated node schemas (#35952) (79a7bd2)
  • core: Respect the editor write lock when Instance AI writes workflows (#36335) (f19c452)
  • core: Return final responses from workflow agents (#36127) (5e2779e)
  • core: Scope AI Assistant credential setup cards to the thread's project (#36425) (b50ea9c)
  • core: Stop auto-selecting the only generic auth credential (#35963) (237c503)
  • core: Stop duplicated workflows from inheriting the original's static data (#35921) (25c1521)
  • core: Stop routing unrelated pre-existing nodes into workflow setup after AI Assistant edit (#36280) (c89dba8)
  • core: Surface model discovery authentication failures to users (#35979) (c44c7de)
  • Default all navigations to domcontentloaded (#36238) (5a1e186)
  • Don't detach child process on Windows so external programs write (#36327) (de6f3d9)
  • editor: Clarify external secret expression previews (#36064) (7792fdb)
  • editor: Display tool execution data in Agent sessions (#36085) (a94ffe9)
  • editor: Improve agent error handoff to AI Assistant (#36076) (598c1f3)
  • editor: Keep agent nodes inside canvas group borders (#35972) (174d0d8)
  • editor: Keep agent tool schemas aligned with saved configuration (#36329) (8d3e266)
  • editor: Keep cursor position when clicking into an expression editor (#36411) (5fa9aaa)
  • editor: Name the AI Assistant browser tab after the conversation (#35958) (e4a414d)
  • editor: Never autosave a read-only preview canvas (#36023) (56a2d20)
  • editor: Only show Markdown placeholder when empty (#36336) (95515d7)
  • editor: Open OAuth sign-in popup before network calls and report blocked pop-ups clearly (#35948) (0f9bff0)
  • editor: Polish workflow header actions menu and production checklist placement (#36209) (f3f075a)
  • editor: Preserve saved credential data after connect-save to prevent false disconnect prompt (#35598) (59ea1f4)
  • editor: Redirect to sign-in when the user's session expires (#35507) (a78ff67)
  • editor: Refresh chat history when switching executions (#35312) (36f2b68)
  • editor: Remove misleading reset zoom control (#35843) (cce5401)
  • editor: Restore the connection test for Kafka and Odoo credentials (#36195) (40a5901)
  • editor: Settle navigation instead of hanging when auth-init guard errors (#36190) (ddd5494)
  • editor: Show agent context after assistant handoff (#36310) (d810315)
  • editor: Show placeholder while MCP settings counts load (#35894) (870099f)
  • editor: Show the connected provider account, not the n8n account email (#35553) (ee03e02)
  • editor: Stop the workflow actions menu shifting when a new workflow is saved (#36077) (a914319)
  • editor: Stop write-lock polling from re-arming after session expiry (#34750) (26be86d)
  • editor: Update ready-to-run demo templates to a current OpenAI model (#36189) (a27c90c)
  • Fix decompress size calculation for zip64 archives (#35856) (d094b6a)
  • Google Ads Node: Migrate from sunset v21 API to v25 (#36255) (bab2649)
  • Google Ads Node: Select the renamed video views metric on API v25 (#36347) (7b64b94)
  • Handle exponential-notation numbers in smartDecimal (#35967) (08052c1)
  • Kafka Trigger Node: Fail v2 activation when the consumer group cannot be joined (#36046) (67a458e)
  • Kafka Trigger Node: Refuse activation when the topic does not exist (#36044) (b025ff0)
  • MCP Server Trigger Node: Keep sessions working when the caller's end-user credentials are unconnected (#36221) (5044810)
  • Microsoft SharePoint Node: Require subdomain in OAuth2 credential (#35680) (620c40c)
  • Microsoft Teams Node: Restore Group.ReadWrite.All OAuth2 scope (#36005) (653dfc6)
  • Microsoft Teams Node: Surface Graph errors instead of returning an empty chat list (#36200) (1cc20d5)
  • Remove default model from one-script docker compose setup (#36034) (a91e012)
  • Rethrow unreachable-element errors in getElementValue (#36199) (1ad64b8)
  • Surface Zoho API error message on duplicate record (#35840) (0a30c03)
  • Wait for task runner registration before executing in e2e test (#36212) (2bca080)
  • Wordpress Node: Send no-cache header on write requests (#20473) (077370f)

Features

  • Adapt the OAuth consent screen for first-party clients (#36261) (2e97fa5)
  • Add LangSmith agent session debug export (#36309) (6cd9a2b)
  • Add snapshot option to every browser use interaction tool (#36111) (7f37137)
  • ai-builder: Add MCP server as allowed tools in skills (#35969) (d0424b4)
  • ai-builder: Auto-select a sensible default model for new agents (#36058) (30aff5d)
  • ai-builder: Offer agent building in assistant greetings (#36073) (e2e719c)
  • API: Add endpoint to create role-mapping rules (#36032) (40a0f2a)
  • API: Add endpoint to get a single role (#36279) (7bd7dac)
  • API: Add endpoint to list all roles (#35609) (48893ee)
  • API: Add endpoint to update a custom role (#36213) (28bf536)
  • API: Migrate workflow tags endpoints to decorator pattern (#35829) (6ab3f5e)
  • Confluence Node: Add Confluence Cloud OAuth2 credential (#35732) (f9fd8a8)
  • Confluence Node: Add page get operation with full sub-tree support (#36217) (d9fa966)
  • core: Add AI Assistant setup funnel telemetry (#36391) (490b368)
  • core: Add an inline Available tools card to connect MCP servers from the assistant chat (#35332) (fb430aa)
  • core: Add env var to skip workflow history compaction on startup (#36099) (5c61888)
  • core: Add get_workflow_versions_diff MCP tool (#35492) (ff7955f)
  • core: Add MCP support for folder creation and workflow moves (#35964) (d4dcc1c)
  • core: Add MCP tool to read data table rows (#35956) (de944d2)
  • core: Add n8n Sandbox support to Agent Knowledge (#35811) (6a30736)
  • core: Add new agent slack integration (#35591) (ca82fd9)
  • core: Add one-off operation flow and insertion-semantics rules to AI Assistant (#36184) (51aa6fb)
  • core: Add writable workspaces to agents (#36033) (01ce449)
  • core: Advertise the MCP 2026-07-28 routing headers in CORS (#35471) (a14ab0b)
  • core: Auto-expose new workflows to MCP from settings via toggle (#35828) (a74fbab)
  • core: Coalesce a Schedule Trigger node's catch-up runs after downtime (#35524) (026cee3)
  • core: Deprecate Array.merge from expressions, and add Array.mergeIntoObject (#36161) (04e76a3)
  • core: Enable Agent sandboxes through Instance AI setup (#36056) (5f9602b)
  • core: Isolate Agent sandboxes by principal (#36203) (ce1b521)
  • core: Let agent builder inspect and update schedules (#36449) (f7459c8)
  • core: Make n8n User Auth (OAuth2) generally available (#35900) (5bfcde5)
  • core: Redesign the account connect panel on hosted forms (#36344) (bf28af6)
  • core: Refuse durable pollers when an active workflow has duplicate trigger node ids (#36340) (faf4dc3)
  • core: Restrict end-user credential creation to team projects (#36025) (e243656)
  • core: Restrict workflow review decisions to assigned reviewers (#36040) (64dcebb)
  • core: Send projectId to the AI Gateway for usage attribution (#35885) (3300833)
  • core: Skip update approval for workflows created in the same Instance AI session (#36054) (dfa841a)
  • core: Support workflow version policies in package exports (#35962) (aecd164)
  • core: Tell the assistant which services are connected and what tools they expose (#35931) (3571c09)
  • Create workflow tools from the agent builder (#36460) (35434b0)
  • Discord Node: Add member moderation actions (#33486) (6bd31e5)
  • editor: Add docked agent preview chat (#35773) (42130c5)
  • editor: Add preview chat to SessionTimeline view (#36128) (dbb28af)
  • editor: Display authors on the review details view (#35897) (d35bb3e)
  • editor: Improve agent session tool call feedback (#36346) (b27a53e)
  • editor: Integrate n8n credits into the node credential picker (#35758) (b2df4c8)
  • editor: Let users update review descriptions (#36069) (cbb5577)
  • editor: Link review success toasts (#35888) (3a4f356)
  • editor: Show character counters on review title and description inputs (#36177) (998fc85)
  • editor: Split workflow review inbox sections (#36216) (dd82601)
  • editor: Switch credits pill after a top-up or depleted allowance (#36239) (516c19e)
  • editor: Update NumberInput styling (#35937) (a319c76)
  • editor: Update SegmentControl component (#34447) (60736dc)
  • Gate Browser Use setup steps on extension detection (#36235) (7025284)
  • Hide n8n-connect usage when aiGatewayCloudUbb is enabled (#35978) (e7fa252)
  • Improve Browser Use connect conditions (#36215) (fee8450)
  • lmChatOpenRouter Node: Add provider routing options (#24822) (d7dcba5)
  • MCP Server Trigger Node: Add server instructions option (#35945) (bfd2746)
  • MiniMax Node: Add M3 and H3 model support (#35935) (aee23f7)
  • Schedule Trigger Node: Add "If Execution Is Missed" option (#35626) (619410f)
  • Schedule Trigger Node: Add a per-node missed execution grace period (#35771) (dc55ed1)
  • Show scopes for managed OAuth credentials with N8N_MANAGED_OAUTH_SHOW_SCOPES (#36053) (5e5bb31)
  • Simplify user flow for Browser Use connection (#36316) (f44bd48)
  • Track AI created credentials success signals (#35812) (8b19044)

Review in cubic

Don't miss a new n8n release

NewReleases is sending notifications on new releases.