AiAdvisors
PL
//AiAdvisors · data updated weekly

n8n AI Automation Index.

How AI agents build automations in n8n. Eight questions answered every week from n8n-mcp telemetry and a corpus of AI-built workflows.

Last updated
Oct 23, 2026
Data window
Oct 5–11, 2026
Author
Romuald Członkowski
80,197
n8n workflows created by AI agents in the week
2.05M
n8n-mcp tool calls in the week
77%
of new workflows contain a Code node
45%
of new installs return the following week

Which LLM powers n8n AI agents: OpenAI, Anthropic, Gemini or Ollama?

In n8n workflows built last week, OpenAI holds 34% of model nodes, Anthropic 21%, Google Gemini 19% and OpenRouter 10%. In Mar 2026 OpenAI had 48% and Anthropic 12%.

Model nodes in workflows built last week31,558 workflows
  1. OpenAI34%
  2. Anthropic21%
  3. Google Gemini19%
  4. OpenRouter10%
  5. Azure OpenAI6%
  6. Ollama4%
  7. DeepSeek2%
  8. Groq1%

Shares are measured among model nodes in workflows that entered the corpus in the period, so they reflect current choices rather than the installed base. Anthropic holds 18% of model nodes in Sep 2026 versus 12% in Mar 2026; Google Gemini 15% versus 16%.

Full answer and history

Which n8n nodes do AI agents use most, and which trigger nodes start their workflows?

The most used node in AI-built n8n workflows is Code, present in 77% of last week's workflows, followed by HTTP Request (59%) and Webhook (49%). The median workflow has 7 nodes.

Share of workflows containing the nodelast week of the corpus
  1. Code77%
  2. HTTP Request59%
  3. Webhook49%
  4. IF38%
  5. Schedule Trigger25%
  6. Manual Trigger19%
  7. Edit Fields (Set)18%
  8. Respond to Webhook16%
  9. Execute Workflow15%
  10. Execute Workflow Trigger14%
  11. Postgres13%
  12. Data Table12%
  13. Google Sheets9%
  14. Telegram7%
  15. Switch7%

The top three are general-purpose tools: code, API calls and webhook entry. Application integrations come lower, led by Google Sheets, Postgres and Telegram. 15% of workflows call another workflow, so composition from sub-workflows is the norm.

Full answer and history

Which n8n node is growing fastest?

The fastest-growing n8n node is stopAndError: its share of AI-built workflows rose from 0.5% in Jun 2026 to 1.7% in Sep 2026, a 3.4× increase. Extract from File is second (2.9×).

Share growth over the last quarterJun 2026 → Sep 2026
  1. stopAndError3.39× · 1.7%
  2. Extract from File2.92× · 2.3%
  3. Data Table2.00× · 9.7%
  4. Convert to File1.48× · 1.2%
  5. Execute Workflow1.40× · 12.5%
  6. Read/Write Files from Disk1.35× · 1.3%
  7. IF1.25× · 35.4%
  8. Structured Output Parser1.25× · 1.6%
  9. Limit1.24× · 1.0%
  10. Wait1.18× · 5.9%

The ranking compares the last complete month with the month a quarter earlier and only counts nodes with at least a 1% share in the last month, so a jump from 0.01% to 0.1% cannot beat real adoption. Searches are a leading signal: an agent searches for a node when it is new or undocumented in its context.

Full answer and history

Do AI-built n8n workflows handle errors?

Rarely. 2.3% of AI-built n8n workflows from last week contain an Error Trigger node, and 2.0% of those built in Sep 2026. Since March the share has never exceeded 2.5%.

Share of workflows with an Error Trigger node, by month
  1. Mar 20261.9%
  2. Apr 20262.5%
  3. May 20262.4%
  4. Jun 20262.2%
  5. Jul 20262.0%
  6. Aug 20262.0%
  7. Sep 20262.0%
  8. Oct 20262.5%

An agent builds the happy path because that is what the user asked for. Retries, failure alerts and a separate error path appear only when someone asks for them explicitly or after the workflow has already failed.

Full answer and history

Do AI agents build n8n workflows from nodes or from code?

From code. The Code node appears in 77% of AI-built n8n workflows from last week and HTTP Request in 59%. In Sep 2026 the Code share was 73%. Agents would rather write JavaScript and call an API than look for a ready-made integration.

Share of workflows with a Code node, by month
  1. Mar 202653%
  2. Apr 202670%
  3. May 202674%
  4. Jun 202671%
  5. Jul 202674%
  6. Aug 202673%
  7. Sep 202673%
  8. Oct 202676%

For the team that inherits these workflows, it means more logic in code and less in configuration. A Code node is harder to review in the editor and does not benefit from the built-in authentication and pagination that native nodes provide.

Full answer and history

How large are AI-built n8n workflows, and which fail validation?

52% of workflows created last week have at least 11 nodes and 19% have 31 or more. The larger the workflow, the less often it fails validation: 1.2% for 2–3 nodes versus 0.15% for 31–100.

Validation failure rate by workflow size80,197 workflows in the week
  1. 2-3 nodes1.18%
  2. 4-10 nodes0.57%
  3. 11-30 nodes0.35%
  4. 31-100 nodes0.15%
  5. 101-300 nodes0.15%

The picture of AI producing only simple three-node automations does not hold. The validation result looks counter-intuitive at first: large workflows are built incrementally through many small updates, each of them checked (7 updates per create). Small workflows are often one-off drafts sent without a check.

Full answer and history

Last week the most searched n8n node was data table (192 installs), ahead of webhook (137) and http request (74). The ranking counts unique installs, not queries.

Searched nodes by number of installsmin. 3 installs per query
  1. data table192
  2. webhook137
  3. http request74
  4. telegram44
  5. google sheets42
  6. schedule trigger41
  7. anthropic37
  8. whatsapp34
  9. slack27
  10. gemini27
  11. mcp client25
  12. ai agent24

Searches reveal what agents do not know by heart: new nodes and application integrations. General-purpose nodes such as Code or IF are almost never searched although they appear in most workflows; messengers and spreadsheets are searched often while ending up in a small share of workflows. The ranking counts unique installs per week with a floor of three installs per query.

Full answer and history

How do AI agents (Claude Code, Cursor) build n8n workflows through MCP: build or maintain?

Maintain. Of 2,054,044 tool calls last week, 61% read workflows, executions and lists, and 11% write. Fetching executions alone is 36% of calls. In the same week 80,197 workflows were created, 11,457 a day.

Workflows created per day
  1. Mon 512,194
  2. Tue 613,540
  3. Wed 713,559
  4. Thu 813,874
  5. Fri 911,936
  6. Sat 107,490
  7. Sun 117,604

Agents check what happened after a run before fixing the workflow. Each create is followed by 7 updates. Only 8% of active installs created a workflow during the week, and 319 installs (2.7% of creators) built 31 or more each.

Full answer and history
//Weekly

What changed this week

A short record of changes between issues. Full answers and charts live on the question pages.

On 27 September 2026 I moved n8n-mcp telemetry to new infrastructure. Versions older than 2.90.0 stopped reporting that day, so figures from before and after it come from different sets of installs. I don't compare one with the other.

//n8n pitfalls

Mistakes AI agents make in n8n

Known n8n pitfalls that n8n-mcp hands to AI agents as context before they build. The list is maintained by hand from mistakes seen in real deployments. One pitfall goes into the weekly issue each week.

Methodology

n8n-mcp telemetry is anonymous and opt-out; it contains no workflow contents or personal data, and install identifiers are hashed. Events are aggregated daily in UTC with roughly an eight-day lag.

The workflow corpus consists of workflows built or modified by AI agents, deduplicated by structure and screened for secrets and personal data. The month is the date a workflow entered the corpus, not its first creation date, which is why we report shares rather than absolute counts.

Searches are counted by unique installs, not by number of queries, so a single looping install cannot dominate the ranking.

Pitfalls are a hand-maintained list of known n8n behaviours that n8n-mcp hands to AI agents as context before building a workflow. They are not derived from user data.

Every figure is generated by a script from the same queries each week; the prose changes together with the numbers.

How to cite

Członkowski, R. (2026). n8n AI Automation Index, week 41/2026 (Oct 5–11, 2026). AiAdvisors. https://aiadvisors.pl/en/n8n-ai-index

Data and charts may be reused with attribution (CC BY 4.0). Source data (JSON)

//Implementation

Need someone to build it?

I build and maintain n8n automations for clients with the same method this index measures: agents build, I own the result, and the team learns to build on its own. First stage from €1,000.

n8n consulting and implementation
//Contact

Let's talk about AI in your business

Tell Alicja what takes up time in your business. She'll book you a free strategy consultation with me.

Talk to Alicja

Prefer a calendar? Pick a time

or write: romuald@aiadvisors.pl · +48 695 263 884