n8n weekly roundup: 5 security fixes, 2.48 beta and Nod

n8n published five security advisories on November 13, two of them High, with fixes in 2.47.6 (stable), 2.48.3 (beta), 3.2.2 (3.x) and 1.123.96 (1.x). The same week, 2.47 became the stable release, 2.48 arrived on the beta track, and n8n Labs released Nod, an open-source personal agent that runs outside n8n. If you self-host, upgrade to 2.47.7, the latest stable patch, which includes the security fixes.

The short version

• Security: five advisories (2 High, 3 Medium). The two High ones are account takeover risks. Cloud is patched automatically.
• Stable: 2.47.7, after 2.47.3 through 2.47.7 between November 9 and 13. It includes Public API rate limits, new in 2.47.0.
• Beta: 2.48.0 (November 10) through 2.48.4 (November 13), with Claude Opus 5.5 as n8n Assistant's default model.
• 3.x beta: 3.2.0, 3.2.1 and 3.2.2, Docker only. Stable 3.x is still planned for early December.
• Nod: a research preview from n8n Labs, a standalone personal agent for Mac and Windows. No n8n account needed.
• Gemini 4: n8n added Gemini 4 Carbon and Gemini 4 Hydrogen from Google.
• Events: n8n at Microsoft Ignite in San Francisco (November 17-19), a Builder Night with Vercel and Arize AI, and hackathons in Vancouver, Santa Cruz and Quito.

Security: five advisories fixed on November 13

n8n's November 13 security update lists five advisories. The two High ones:

• The invitation acceptance endpoint lets a member take over pending user accounts.
• A custom instance role with user management permissions gives control of the owner account.

The three Medium ones cover workflow activation letting trigger nodes use credentials the publisher can't access, AWS Bedrock nodes skipping SSRF protection for custom runtime endpoints, and integration nodes sending authenticated requests to unintended API endpoints.

The fixed versions are 2.47.6 (stable), 2.48.3 (beta), 3.2.2 (3.x) and 1.123.96 (1.x). n8n says Cloud instances are patched automatically, and recommends that self-hosters on older versions upgrade. Instances with several users matter most here, because both High advisories are about accounts. Our security update post goes through each advisory and who is affected.

n8n 2.47 is now the stable release

On November 14, npm's stable and latest tags point to 2.47.7 (npm). The 2.47 line got five patches this week, per n8n's changelog:

• 2.47.3 (November 9): only users who can edit a workflow can own its Migration Report findings, and notification emails open the report filtered to the recipient.
• 2.47.4 (November 11) and 2.47.5 (November 12): expression and editor fixes, and the Public API now rejects edits to read-only credential nodes.
• 2.47.6 (November 13): the security fixes, plus a Pinecone Vector Store fix and a faster expression runtime.
• 2.47.7 (November 13): one fix, "Stop redacting executions that use global credentials" (release). If you already moved to 2.47.6, take this one too.

2.47 brings Public API rate limits to stable. n8n's blog post says they're available from 2.47.0 on every plan, Cloud and self-hosted, and off by default. Admins turn them on in Settings > Policies > Public API Limits, through the API, or on self-hosted with N8N_PUBLIC_API_RATE_LIMIT_ENABLED=true. Limits count per API key, clients over the limit get a 429 with Retry-After, and requests from the editor don't count. Workflows that call their own instance's API should use the node's Batching option. Our guide to rate limits, 429 errors and batching covers that pattern in general.

If you're coming from 2.46, our 2.47 roundup covers the rest of the release. The 2.46 line got one more patch, 2.46.7 (November 9), which limits resume attempts for waiting and queued executions.

n8n 2.48 on the beta track

n8n 2.48.0 came out on November 10, and 2.48.4 on November 13 is now on npm's beta and next tags. From the 2.48.0 changelog:

• n8n Assistant's default model is now Claude Opus 5.5.
• The redaction policy now applies to log streaming, telemetry and error reports.
• The Migration Report hides archived workflows by default, shows a message when it's slow to load, and opens filtered to the recipient from notification emails.
• A Microsoft Work IQ OAuth2 credential, and agent runs stop when the Gateway credits wallet is empty.
• The Paddle node is renamed to Paddle Classic.

2.48.1 to 2.48.4 carry the same fixes as this week's 2.47 patches, with the security fixes in 2.48.3. Our 2.48 roundup has the details and our upgrade test.

n8n 3.x: 3.2.2 carries the security fixes

On Docker Hub, the 3.x beta moved from 3.1.0 to 3.2.0 on November 10, then 3.2.1 and 3.2.2 on November 13. As with every 3.x release so far, there's no npm package. The 3.2.2 changelog lists what it adds on top of 2.48.0, including AI Agent node fixes such as saving every tool-calling round of a turn to chat memory. 3.2.2 is the 3.x version with the security fixes.

The plan hasn't changed: n8n's 3.0 announcement says stable 3.x follows in early December, after which 2.x gets security and critical bug fixes for 12 months. Support for 1.x ends on December 31, 2026, and there's no direct path from 1.x to 3.x. The 1.x line got 1.123.93 through 1.123.96 this week.

Nod: an open-source personal agent from n8n Labs

On November 10, n8n introduced Nod, a research preview from n8n Labs, a new team that experiments with ideas outside the core product. Nod is a standalone app for Mac and Windows, separate from n8n and from the agents you build in n8n, and you don't need an n8n account to use it. It works with its own browser inside a local sandbox, uses the AI provider you connect, and is open source under Apache 2.0.

n8n says Nod doesn't change n8n's direction, and work on n8n and its agent continues as planned. If you use n8n, you can connect your instance so Nod can work with your workflows. Our Nod post covers what it does, how it uses n8n, and its limits.

Gemini 4 Carbon and Hydrogen in n8n

n8n added Gemini 4 Carbon on November 10, which Google calls its strongest model yet. You pick it in the Google Gemini Chat Model of an AI Agent node, or in the standalone Google Gemini node. On n8n Cloud it works with Gateway credits, and on self-hosted with your own Gemini API key. n8n's post quotes Google's introductory price of $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 after November 30, 2026.

Two days later, n8n announced on X that Gemini 4 Hydrogen is in n8n too, a fast model built for tool-use loops and agentic coding. n8n suggests letting Carbon or Argon plan and Hydrogen do the legwork.

From the n8n blog

• The future of work is built by everyone (November 12). The next post in n8n's series on how it runs automation internally argues that everyone in a company should be able to automate their own work. In n8n's summary on X, its automation team became a guild that sets standards and owns critical systems.
• Batch processing patterns for workflow automation at scale (November 10). Patterns for jobs that work at 200 records but stop at 20,000: memory limits, API quotas, and knowing what finished. Our Loop Over Items guide covers the n8n side.
• Microservices best practices (November 10). Service boundaries and failure handling, so one failed API doesn't take everything down with it.
• E-signatures with TurboDocx (November 8). A Verified Node Spotlight guest post by TurboDocx's CEO: generate a contract, approve it in Slack, send it for signature and file the signed PDF in SharePoint.

n8n also posted on X that 24 sessions from In The Loop, its conference last month, are now on its YouTube channel.

n8n events coming up

The notable events from n8n's posts on X and its community calendar. Times are local.

• Microsoft Ignite, San Francisco, November 17-19. n8n is at booth 5344R. On November 19 it plans a live build showing how Microsoft Agent 365 and n8n cover identity, access and orchestration for agents, and the booth runs The n8n Detective Agency, a game where you question suspects through an AI detective built in n8n.
• Enterprise dinner, San Francisco, November 16. n8n posted on X that it's hosting an invite-only dinner in the Financial District with the n8n and Microsoft teams, and that a few spots opened. Requests need approval.
• Builder Night San Francisco: n8n x Vercel x Arize AI, November 17, 5-8 PM. Short, demo-led talks on taking agents from idea to production, including "Agents that run your n8n workflows" by Liam Hampton of n8n. Space is limited to about 60 guests, and n8n's latest post points to a waitlist.
• n8n Meetup San Francisco: Microsoft Ignite Afterhours, November 19, 6-9 PM, at Digital Jungle SF.
• LOOP: UBC eProjects x n8n Buildathon, Vancouver, November 14-15. A two-day AI and automation hackathon with a hands-on n8n workshop, open to beginners and people without a team.
• Nodo Cero: Hackathon n8n Bolivia, Santa Cruz de la Sierra, November 21. The Bolivian n8n community's hackathon with Women Techmakers Santa Cruz, aimed at getting more women building automations and AI agents.
• n8n Hackathon Quito: Build with AI: EduTech, November 27-28. A free 21-hour hackathon where teams build tools for real problems in education. Registration has two steps, explained on the event page.

Everything else is on the n8n Community Events calendar.

In the community forum

Community Weekly Challenge 09: The Balance Check asks which of eight invoices still need payment and how much is outstanding, built with native n8n nodes (no Code node). It closes on Sunday, November 15, at 18:29 UTC.

Topics that drew discussion this week:

• A retry re-sent 49 emails that had already gone out, while Retry On Fail marked the run as a success.
• The workflow failed, but the email still went out: how do you test for that?
• The MCP menu not showing after setting N8N_MCP_MANAGED_BY_ENV=false in 2.47.5.

Also on then8nblog this week

• n8n security update: 5 advisories fixed in 2.47.6 and 2.48.3
• n8n rate limits: handle 429 errors, Retry-After and batching
• Nod: n8n Labs' open-source personal agent, and how it uses n8n
• n8n 2.48: what's new and should you upgrade
• n8n environment variables for production: what to set and why
• n8n Loop Over Items: batches, pagination and the done output
• n8n webhooks: test vs production URL, auth and responses

All our weekly roundups are on the weekly roundup tag page.