n8n Changelog Explained: What Shipped in 2026 and Why It Matters

The n8n changelog is more than a list of version numbers. For the technical teams who build automation on the platform, it is the fastest way to understand which new capabilities are worth adopting, how they change the way workflows run, and what they mean for the way you manage an instance day to day.

This guide walks through the most important n8n changes released in 2026, translating each entry into practical guidance for people who run production automation. Whether you are a solo builder, an in-house platform team, or an AI consultant advising clients on workflow automation, the goal is the same: understand what changed, decide whether it affects you, and act on it.

How to Read the n8n Changelog

Before diving into individual releases, it helps to understand how n8n documents change. The platform maintains several release resources, and each serves a different purpose. Knowing which one to consult saves time and prevents confusion.

The Four Release Resources

• Changelog (this page): A curated, narrative summary of the most important new features as they roll out. It is written to stand on its own, so a single entry is easy to share with a colleague who is waiting on one specific update.
• Release notes: A listing of all feature-level updates in each release. Use this when you need completeness rather than narrative.
• GitHub releases: Full change detail for each release, linked to commits, including bug fixes and minor changes. This is the source of truth for engineers who need to trace a change to its code.
• Archived pages: Old-style release notes pages for the 2.x, 1.x, and 0.x lines remain archived. Everything in the 2.x archive is covered by the current changelog and the Release notes.

For major version upgrades, n8n provides dedicated migration guidance, including v3.0 breaking changes, v2.0 breaking changes, the v2.0 migration tool, and the v1.0 migration guide. If you are planning a version jump, read the breaking-changes documentation before you touch a production instance.

Stable vs. Beta Versions

n8n releases a new minor version most weeks. Two channels matter in practice:

• Stable: The version intended for production use.
• Beta: The most recent release. It may be unstable.

At the time of writing, the current stable version is 2.39.5 and the current beta is 2.40.0. A simple operational rule follows from this: pin production instances to stable, and test beta in a staging environment if you want early visibility into upcoming changes. To report issues, the n8n community forum is the recommended channel.

Error Workflow Executions No Longer Count Toward Your Quota

Released: 2026-09-01 in n8n 2.38

This is one of those changes that sounds small and turns out to be significant for anyone running automation at scale.

What Changed

Runs of your error workflows are now excluded from your execution quota, on every plan. An error workflow is the one n8n runs when another workflow fails. It receives the details of the failure – which workflow failed, which node, and what went wrong – and does something useful with them: post to a Slack channel, send an email, open a ticket, or retry the run.

Until this release, each of those runs counted like any other execution. Handling a failure cost you an execution on top of the failure itself. That created a perverse incentive: some teams left error workflows out entirely to protect their quota, which meant failures went unnoticed until someone downstream complained.

Why It Matters

Every production workflow should have an error workflow attached. The quota change removes the last practical reason to skip one. If you have been running workflows without error handling because of execution limits, this release is your cue to fix that.

How to Take Advantage

1. Open the workflow you want to protect and go to Workflow Settings.
2. Attach an error workflow. Its runs are then excluded from the count.
3. One error workflow can serve as many workflows as you like, so you can build a single, well-tested handler and reuse it across your entire instance.

On n8n Cloud, the change is already live. On self-hosted deployments, it applies from 2.38.0, or 1.123.60 if you are still on the v1 line.

Practical example: A data team runs a nightly ETL workflow that pulls from three APIs. An error workflow catches failures, posts the failing node and error message to Slack, and opens a ticket in their project tracker. Before 2.38, a bad night could consume dozens of executions just on error handling. Now those runs are free, and the team no longer hesitates to add error workflows to every new automation.

If some of your workflows still run without an error workflow, refer to n8n's "Handle errors gracefully" documentation to set one up.

Use AI Models and Tool Services Without Provider Accounts or Credentials

Released: 2026-09-02 in n8n 2.36

Setting up AI in an automation platform has traditionally meant a scavenger hunt: create an account with each provider, add a payment method, generate an API key, and store it as a credential. Gateway credits remove that friction.

What Gateway Credits Do

You can now use supported AI models and services in n8n Cloud without first creating an account with each provider, setting up billing, or adding an API key. Gateway credits let you start building with a prepaid balance managed through n8n.

Supported AI providers include:

• OpenAI
• Anthropic
• Gemini directory on Neura Market
• Alibaba Cloud Qwen
• MiniMax
• Moonshot Kimi

Credits also work with several tool services: Brave Search, Firecrawl, Browserbase, LlamaParse, and PDF.co. That combination matters for agent-style workflows, where a model reasons over results returned by search, scraping, or document parsing services.

How to Use Them

On a supported node, select Gateway credits when setting up the credential, then run your workflow. The choice is made per node, which means the same workflow can use Gateway credits for one service and your own provider credentials for another. That flexibility is useful during migration or when a specific service has negotiated pricing you want to keep.

Usage is deducted from a shared prepaid balance for the n8n instance. n8n aligns Gateway credit rates with publicly listed provider pricing wherever possible and publishes the rates for every supported service on its service pricing page.

Managing the Balance

Instance owners can add funds manually or configure automatic top-ups based on a balance threshold, a target balance, and an optional monthly limit. You can track remaining balance and spend from Cloud Admin Panel > Manage > Gateway credits, including breakdowns by model and workflow.

Practical example: An AI consultant prototyping a document-processing workflow for a client can wire up Anthropic, LlamaParse, and Brave Search in a single afternoon without asking the client's IT team for three separate vendor accounts. Once the prototype proves out, the consultant can swap individual nodes to the client's own credentials without rebuilding the workflow.

You can continue using your own provider credentials as before. Gateway credits are an additional option, not a replacement. For more detail, see the n8n forum post announcing the feature.

n8n Assistant on Self-Hosted: Set Up in Minutes

Released: 2026-08-18 in n8n 2.35

n8n Assistant arrived on n8n Cloud in July and has worked on self-hosted n8n since then. The catch was the setup: getting there meant enabling the instance-ai module and configuring a sandbox, a model, and web search through environment variables.

n8n 2.35 changes that in three ways: it enables the module by default, adds a one-line install that pre-configures the pieces you would otherwise assemble yourself, and allows you to choose your model provider, sandbox, and search directly in the UI.

The Three Components

Self-hosted setup needs two things, plus one worth adding:

1. A model provider. Bring an API key for Anthropic, OpenAI, or OpenRouter, or point n8n at any OpenAI-compatible endpoint, including a local one. You pay the provider directly, and your prompts, workflow content, and the execution data the assistant reads go to that provider. That last point matters for teams with data residency or confidentiality requirements – review it before enabling the assistant on sensitive workflows.

2. A sandbox. The assistant runs code to build and test what you ask for, and that code executes in an isolated sandbox, never on your n8n server. n8n's bundled sandbox runs on your own Docker host and suits local development and testing. For production, n8n recommends Daytona's managed sandbox with an API key.

3. Web search (optional but valuable). With it, the assistant reads current docs and APIs instead of relying on what its model remembers. It works through a bundled SearXNG instance or a Brave Search API key. Given how quickly APIs change, this is the difference between an assistant that suggests a deprecated endpoint and one that suggests a working one.

Installation Paths

On a brand-new instance, one command sets up everything except the model key. It installs n8n with Docker Compose and pre-configures the sandbox and SearXNG web search, both free. Open the editor, add your model API key in the instance's AI settings, and start building.

On an existing Docker install, the n8n Assistant setup guide covers the sandbox and search options for each deployment shape, in environment variables or in the UI.

Important Constraints

• n8n Assistant requires Docker. npm installs will stop working with n8n 3.0 in October, so new installs should use the one-line setup or Docker Compose.
• n8n Assistant is in Preview. It can make mistakes, and its behavior may change while it is in development. Always review generated workflows before using them in production.
• Self-hosted Enterprise support is coming. Enterprise customers who want to try n8n Assistant before then can contact their Customer Success Manager about preview access.

Practical example: A platform team at a mid-size company uses the one-line install to stand up a test instance, points the assistant at a local OpenAI-compatible model for cost control, and uses it to draft boilerplate workflows. Because the sandbox is isolated, they can let the assistant execute test code without exposing their production server.

Return Webhook Responses of Any Size from Your Workers

Released: 2026-08-04 in n8n 2.34

This release solves a specific but painful limitation in queue mode deployments.

The Problem

In queue mode, the worker runs the execution, but the client that made the request stays connected to the main or webhook instance. A response from a Respond to Webhook node therefore has to travel back through the queue. Redis holds the whole response while that message is in flight, which is why n8n caps it at 64 MiB by default (the N8N_WEBHOOK_RESPONSE_RELAY_SIZE_MAX setting).

A response that serialized to more than that failed the node. Common culprits included a wide result set from a database query, a batch of aggregated API calls, or a CSV or XML document assembled in the workflow.

The Fix

Set N8N_WEBHOOK_RESPONSE_RELAY_OFFLOAD_ENABLED=true on your workers. n8n then stores a response body above the limit in binary data storage instead. The queue message carries only a reference, the main instance streams the body from storage to the client, and n8n deletes the stored copy once it delivers the response, so nothing accumulates. Redis memory stays flat no matter how large the response gets.

Requirements and Caveats

Offloading needs a N8N_DEFAULT_BINARY_DATA_MODE that stores data (any mode except default) and storage that every instance can read. n8n recommends s3 or azure, since both stream the body.

Practical example: An integration team exposes a webhook that returns a generated CSV report of the previous day's orders. The report can easily exceed 64 MiB during peak season. With offloading enabled and S3 configured, the worker writes the CSV to object storage, the main instance streams it to the client, and the stored copy is deleted after delivery. No more failed nodes during the busiest weeks of the year.

Putting It Together: A Practical Adoption Plan

The changes above span four releases, and adopting all of them at once is rarely the right move. A sensible sequence for most teams looks like this:

1. Audit error workflows first. This is the highest-value, lowest-risk change. Inventory your production workflows, identify which ones lack an error workflow, and attach one. On Cloud, the quota benefit is already live; on self-hosted, upgrade to 2.38.0 or later (or 1.123.60 on v1).
2. Evaluate Gateway credits for prototyping. If your team has been slow to experiment with AI nodes because of provider account overhead, Gateway credits remove that barrier. Start with a small prepaid balance and track spend in the Cloud Admin Panel.
3. Trial n8n Assistant in a non-production instance. Because it is in Preview, treat it as a productivity tool for drafting and testing, not as an unattended production component. Review every generated workflow before it ships.
4. Enable webhook response offloading only if you have hit the limit. This is a targeted fix. If your queue mode deployments never return large payloads, you do not need it. If they do, configure S3 or Azure storage and set the environment variable on your workers.

For teams advising clients on automation strategy, the pattern is worth noting: n8n is steadily reducing the operational friction of running automation at scale – quota mechanics, credential management, setup complexity, and infrastructure limits. Each release removes a reason to avoid doing the right thing, whether that is adding error handling or testing AI-assisted workflow generation.

Frequently Asked Questions

Does the error workflow quota change apply to self-hosted instances? Yes. On self-hosted it applies from 2.38.0, or 1.123.60 if you are on the v1 line. On Cloud it is already live.

Can I mix Gateway credits and my own provider credentials in one workflow? Yes. The choice is made per node, so a single workflow can use Gateway credits for one service and your own credentials for another.

Where do my prompts and execution data go when I use a self-hosted model provider with n8n Assistant? They go to the provider you configure. You pay that provider directly, so review your data handling requirements before enabling the assistant on sensitive workflows.

Is n8n Assistant production-ready? It is in Preview. It can make mistakes, and its behavior may change. Review generated workflows before using them in production.

What happens to npm installs with n8n 3.0? npm installs will stop working with n8n 3.0 in October. New installs should use the one-line setup or Docker Compose.

Keeping Up With Future Releases

n8n ships a new minor version most weeks, which means the changelog is a living document rather than a one-time read. A lightweight routine keeps you current without drowning in release notes:

• Subscribe to the changelog and skim the narrative entries. They are written to stand alone, so you can triage quickly.
• When an entry affects you, follow the link to the Release notes for the full feature list and to GitHub releases for commit-level detail.
• Before any major version upgrade, read the breaking-changes documentation for that version and test in staging.
• Pin production to stable and use beta only in environments where instability is acceptable.

Used this way, the changelog becomes less of a documentation page and more of an operating rhythm – a regular checkpoint that keeps your workflows, your instance, and your team aligned with where the platform is heading.

Frequently Asked Questions

What is the best way to get started with n8n Changelog Explained: 2026 Releases, ?

The best approach is to start with a clear goal in mind. Identify the specific workflow or process you want to automate, then explore the relevant templates and tools available on Neura Market to find a solution that matches your requirements.

How much does workflow automation typically cost?

Costs vary significantly depending on the platform and scale. Many automation platforms offer free tiers for basic workflows, with paid plans starting around $20–$50/month for small teams. Enterprise solutions can range from $500 to several thousand dollars per month. Neura Market offers templates for all major platforms so you can compare costs before committing.

Do I need technical skills to implement workflow automation?

Modern no-code and low-code platforms like Zapier, Make.com, and others have made automation accessible to non-technical users. Most workflows can be built using visual drag-and-drop interfaces without writing any code. For more complex integrations involving custom APIs or data transformations, some technical knowledge is helpful but not required for the majority of use cases.