What n8n-io/n8n shipped
Written by FoxPlug from the repository's own releases; not affiliated with n8n. Paste your own repository in the box below and FoxPlug writes the same weekly update for your product. This page is an automatic summary of the releases, pull requests and commits in github.com/n8n-io/n8n; n8n did not write it and does not use or endorse FoxPlug. Every line links to the change it describes.
Get a weekly update like this for your product, free
Week of October 5, 2026
What shipped
- Assistant home now offers User and Power modes, with Power mode showing workflows, agents, approvals, people and the n8n MCP boundary. Pull request #40838
- Model providers restricted by policy are now disabled in the agent model selector with a lock icon and explanation. Pull request #40658
- Vertex AI credentials now support project selection from a dropdown or manual entry of a custom Project ID. Pull request #40667
- Migration rule details now include quick filters, sorting, and a filters popover to narrow results by status and metadata. Pull request #40781
- External agent channels now receive a fixed message when an agent run stops due to budget guardrails on Slack, Telegram, Discord, Teams, WhatsApp and Linear. Pull request #40767
- Public API controller routes now support binary response bodies with media type and header declarations. Pull request #40704
- Inactive frontend modules can now declare a placeholder page to show a paywall for licensed features. Pull request #40553
- Workflow and node executions now emit opt-in marker spans with zero duration to signal execution start before completion. Pull request #40762
- Agent chat messages now send optional client context like timezone from the user's browser. Pull request #40612
- WhatsApp inbound media handling now includes images, documents, audio, voice notes, video, stickers, locations and shared contacts. Pull request #39409
Why it matters
This week adds polish to agent features and policy controls while expanding credential options and improving workflow tracing. The assistant gains distinct user and power modes, model provider restrictions are now enforced in the UI, and external channels get clearer feedback on budget stops. These changes make the agent system more controllable and the assistant more discoverable for different user types.
Changelog entry
- Feat(editor): Assistant home now offers User mode for basic queries and Power mode showing workflows, agents, approvals, people and n8n MCP boundary, with browser preference persistence Pull request #40838
- Feat(editor): Model providers restricted by policy are disabled in agent model selector with lock icon and scope explanation Pull request #40658
- Feat(editor): Vertex AI credentials support project selection from dropdown or custom Project ID entry Pull request #40667
- Feat(editor): Migration rule details include quick filters, sort menu and filters popover with status and metadata options Pull request #40781
- Feat(core): External agent channels receive budget-stop message on Slack, Telegram, Discord, Teams, WhatsApp and Linear Pull request #40767
- Feat(API): Public API controller routes support binary response bodies with media type and header declarations Pull request #40704
- Feat(editor): Inactive frontend modules can declare placeholder page for licensed features Pull request #40553
- Feat(core): Workflow and node executions emit opt-in marker spans with zero duration to signal execution start Pull request #40762
- Feat(core): Agent chat messages send optional client context like timezone from user browser Pull request #40612
- Feat(core): WhatsApp inbound media handling covers images, documents, audio, voice notes, video, stickers, locations and contacts Pull request #39409
- Feat(core): Credential type policies support extends selector to match credential types built on a base type at any depth Pull request #40305
- Feat(core): Deployment artifact now reported in telemetry to track installation methods and versions Pull request #40319
This week: Assistant home splits into User and Power modes, model providers respect policies, Vertex AI project selection, and agent budget-stop messages reach external channels.
This week in n8n: The Assistant home now separates User and Power modes for different workflows. Model provider restrictions are enforced in the agent selector. Vertex AI credentials support project selection. External channels receive clear messages when agent runs hit budget limits. Migration tools gain filtering and sorting. These updates improve policy enforcement, credential management, and user guidance across the platform.
Week of September 28, 2026
What shipped
- Administrators can now toggle Agents on or off in instance settings, with the choice persisting across restarts and license changes. Pull request #40153
- N8n Chat is now available as an agent channel, allowing builders to make agents available in chat with custom descriptions. Pull request #40075
- Workflow history pruning now runs as a durable system task, so in multi-main setups only one process executes the cleanup instead of all. Pull request #40106
- The control plane can now stop engine v2 executions and deliver cancellation lifecycle events. Pull request #39786
- Preview stop requests now pause background sub-agents at safe checkpoints and send a combined report after all selected jobs become inactive. Pull request #40119
- The migration rule detail page now reads from the
migration_findingtable instead of scanning every workflow, improving performance. Pull request #39974 - Text input heights on the project settings page are now aligned to a consistent row height. Pull request #39611
- Node icons that declare sizes smaller than their slot now scale up to fill the available space on the canvas. Pull request #40181
- Assistant thinking duration no longer includes the time span added when restarting during a run and marking tool calls as interrupted. Pull request #40161
- LDAP configuration endpoints migrated to PublicApiController with generated OpenAPI documentation. Pull request #39966
Why it matters
This week brings agent management improvements with instance-level toggles and n8n Chat integration, performance optimizations for multi-process setups and large migrations, and better reliability in execution control and UI consistency. These changes support production deployments at scale while improving the agent builder experience.
Changelog entry
- Instance setting to toggle Agents on or off under Settings > Agents, persisting across restarts Pull request #40153
- N8n Chat added as an agent channel with builder descriptions (125_agents_n8n_chat feature flag) Pull request #40075
- Workflow history pruning now executes as durable system task in multi-main setups Pull request #40106
- Control plane can now stop engine v2 executions with cancellation lifecycle events Pull request #39786
- Preview pause requests now coordinate background sub-agents and report after all jobs inactive Pull request #40119
- Migration rule detail page loads from
migration_findingtable instead of full workflow scan Pull request #39974 - Project settings text inputs aligned to consistent row height Pull request #39611
- Node icons now scale up to fill slot space when declared size is smaller Pull request #40181
- Assistant thinking duration excludes restart sweep time from interrupted tool calls Pull request #40161
- ProjectScopeService moved to @n8n/backend-services for shared usage Pull request #40209
- WorkflowSharingService moved to @n8n/backend-services with improved repository methods Pull request #40157
- LDAP configuration endpoints migrated to PublicApiController with OpenAPI generation Pull request #39966
N8n now supports agents in n8n Chat, adds instance-level agent toggles, improves multi-process efficiency, and optimizes migration performance.
This week's n8n releases include three key improvements: agents are now available as a channel in n8n Chat with admin control toggles, workflow history pruning runs more efficiently in multi-process setups, and migration rule pages load faster by reading from indexed tables. These changes strengthen agent deployment capabilities and production reliability.
Week of September 21, 2026
What shipped
- Added database foundation for a durable agent message queue that accepts messages in FIFO order with session locking across execution. Pull request #39435
- Connected Preview and integration messages to the agent queue so busy sessions keep new messages pending with independent runs. Pull request #39468
- Let users create missing credentials and variables directly from the promotion binding screen with prefilled names and types. Pull request #39342
- Workflow save now accepts new group shapes including groups that hold triggers together with nodes. Pull request #39236
- Engine v2 executions now report waiting status when every step is suspended instead of showing running. Pull request #39093
- Scheduled, polled, and production webhook executions now record the user who published the workflow version. Pull request #39466
- Canvas toolbar buttons are now grouped in a single container to save space on artifact and editor pages. Pull request #39587
- Added package import-selection command to @n8n/cli tool to import a subset of workflows from an.n8np package. Pull request #39566
- Agent skills modal now starts with file or folder upload and manual entry options with drag-and-drop support. Pull request #39545
Why it matters
Agent queueing and message handling received foundational improvements for better concurrency. Workflow promotion and credential management got easier with binding-time creation. Several editor and core execution features improve visibility and functionality for running workflows.
Changelog entry
- Feat(core): Add the durable agent message queue foundation with database entity, migration, and repo Pull request #39435
- Feat(core): Connect Preview and integrations to the agent queue with independent runs per message Pull request #39468
- Feat(editor): Create credentials and variables from promotion bindings with prefilled names and types Pull request #39342
- Feat(core): Accept new group shapes when workflow is saved including trigger-containing groups Pull request #39236
- Feat(engine): Store execution waiting status when every step is suspended Pull request #39093
- Feat(core): Attach publishing user to triggered executions from schedules, polls, and webhooks Pull request #39466
- Feat(editor): Group canvas toolbar buttons in single container with toolbar variant Pull request #39587
- Feat: Add package import-selection command to @n8n/cli for subset workflow imports Pull request #39566
- Feat(editor): Revamp Agent skills modal with file upload and manual entry options Pull request #39545
- Fix(engine): Handle Redis failure modes in engine v2 execution responses Pull request #39486
- Refactor(core): Encapsulate webhook response state in WebhookResponder class Pull request #39626
- Feat(editor): Add telemetry for agent channel setup funnel Pull request #39604
- Fix(Google Cloud Storage Node): Preserve request headers on object create with Service Account Pull request #35218
- Fix(editor): Show community package updates when only verified packages enabled Pull request #39424
- Fix(Microsoft Teams Node): Correct v2 resource-locator copy strings Pull request #39443
- Fix(core): Accept nested folders when applying promotion package Pull request #39630
- Feat(engine): Support binary reference bodies in Engine v2 sendResponse Pull request #39534
This week: durable agent message queues, in-editor credential creation from promotions, new group shapes for workflows, and waiting status visibility in engine v2.
This week in n8n: we added durable agent message queue foundations for handling messages in FIFO order across sessions, let users create credentials and variables directly from promotion bindings, brought new group shapes to workflows including trigger-containing groups, and added waiting status visibility to engine v2 executions when all steps are suspended.