Week 42 / 2026.
A short record of changes between issues. Full answers and charts live on the question pages.
- Data window
- Oct 12–18, 2026
- Last updated
- Oct 30, 2026
- Author
- Romuald Członkowski
Commentary
Between 12 and 18 October, n8n-mcp produced 81,909 workflows, 11,701 a day on average, across 2,005,999 tool calls with a 0.15% error rate. Little moved week on week: 2% more workflows a day, 2% fewer calls, and 151,797 active installs, which is flat. The busiest day was 14 October with 16,078 workflows, the quietest 17 October with 6,795. Week-one retention stands at 39% of the 5,868 installs that started in the week of 5 October, against 45% in the previous issue, and the data does not show what sits behind the difference.
Validation turned away 0.6% of workflows, and it is the smallest ones that trip over it most. Workflows with 2–3 nodes fail validation in 0.98% of cases, those with 4–10 in 0.55%, those with 11–30 in 0.36%, and those with 101–300 nodes in 0.10%. In this data, the size of a workflow does not raise the risk. I'd point at another figure: 13% of connection tests end in an error, and 23% of those come down to an address or the network. That is the cost of configuration, not of building logic.
An Error Trigger appears in 2.8% of the workflows that entered the corpus in this window, and 14% call a sub-workflow. This issue's pitfall sits in exactly that spot: an Error Trigger has to live in a separate workflow, because it will not catch an error in its own. Price error handling as a second workflow, then, not as a node added to the first.
The window added 29,053 workflows to the corpus, which now holds 1,219,933, with a median of 7 nodes. Code appears in 75% of them, HTTP Request in 60%, Data Table in 12.7%, an LLM model node in 6% and an AI Agent in 4%. The model-node share has been falling since March, when it stood at 17%, while Data Table climbed over the same period from 3.3% to 12.4% in the October data. Among model nodes, OpenAI holds 40%, Anthropic 23% and Google Gemini 13%.
Changes versus the previous issue
| figure | value | change |
|---|---|---|
| Workflows created | 81,909 | — |
| Workflows per day | 11,701 | +2% |
| Tool calls | 2,005,999 | -2% |
| Calls ending in error | 0.15% | +48% |
| Failed validation | 0.6% | -10% |
| Active installs | 151,797 | +0% |
| Week-1 retention | 38.8% | -14% |
| Workflows added to corpus (week) | 29,053 | -8% |
This week's answers
- Which LLM powers n8n AI agents: OpenAI, Anthropic, Gemini or Ollama?
In n8n workflows built last week, OpenAI holds 40% of model nodes, Anthropic 23%, Google Gemini 13% and OpenRouter 10%. In Mar 2026 OpenAI had 48% and Anthropic 12%.
- 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 75% of last week's workflows, followed by HTTP Request (60%) and Webhook (48%). The median workflow has 7 nodes.
- 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×).
- Do AI-built n8n workflows handle errors?
Rarely. 2.8% 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%.
- Do AI agents build n8n workflows from nodes or from code?
From code. The Code node appears in 75% of AI-built n8n workflows from last week and HTTP Request in 60%. In Sep 2026 the Code share was 73%. Agents would rather write JavaScript and call an API than look for a ready-made integration.
- How large are AI-built n8n workflows, and which fail validation?
52% of workflows created last week have at least 11 nodes and 18% have 31 or more. The larger the workflow, the less often it fails validation: 1.0% for 2–3 nodes versus 0.14% for 31–100.
- Which n8n nodes do AI agents search for?
Last week the most searched n8n node was data table (225 installs), ahead of webhook (133) and http request (80). The ranking counts unique installs, not queries.
- How do AI agents (Claude Code, Cursor) build n8n workflows through MCP: build or maintain?
Maintain. Of 2,005,999 tool calls last week, 60% read workflows, executions and lists, and 11% write. Fetching executions alone is 37% of calls. In the same week 81,909 workflows were created, 11,701 a day.
Most searched nodes
data table (225) · Webhook (133) · http request (80) · Telegram (53) · Gmail (51) · anthropic (43) · whatsapp (39) · schedule trigger (39) · Slack (38) · gemini (25)
Error Trigger workflow must be separate
The Error Trigger node must be in a SEPARATE workflow from the one being monitored. It cannot catch errors in its own workflow. Create a dedicated error-handling workflow with an Error Trigger, then set each production workflow to use it via Settings > Error Workflow.