Week 41 / 2026.
A short record of changes between issues. Full answers and charts live on the question pages.
- Data window
- Oct 5–11, 2026
- Last updated
- Oct 23, 2026
- Author
- Romuald Członkowski
Commentary
Between 5 and 11 October, n8n-mcp produced 80,197 workflows, 11,457 a day on average, across 2,054,044 tool calls with a 0.10% error rate. That is 10% more workflows a day and 22% more calls than the previous issue, and the data does not show what sits behind the move. There are 151,238 active installs, 8% of them create workflows, and 319 installs hold 31 workflows or more, 2.7% of the creators.
This is the first retention figure after the telemetry move. Of the 6,965 installs that started in the week of 28 September, 45% were still active a week later, which is this window. That cohort's week one covers the whole window, so this is not a reading taken mid-week. I don't hold the figure against issues from before 27 September.
Each workflow created draws 7.3 updates. The window holds 140,248 edits, 2.67 operations and 1.45 changed properties each on average, a median of 877 ms, a verifiable improvement in 56% of them. Among the edits with a recognised intent, changing configuration leads (52,448), adding functionality is more than twice as rare (19,602), and fixing validation accounts for 4,394. Reads make up 61% of tool calls, writes 11%, and fetching executions alone 36% of everything. Budget an implementation for seven returns to a workflow, not for building it once.
This window added 31,558 workflows to the corpus, which now holds 1,190,875, with a median of 7 nodes. Code appears in 77% of them, HTTP Request in 59%, an LLM model node in 6%, an AI Agent in 4%; among model nodes, OpenAI holds 34%, Anthropic 21% and Google Gemini 19%. Data Table is in 12.1% of this window's workflows and has been climbing since March, when it stood at 3.3%; in the October data it sits at 12.2%. "data table" was also the most searched phrase, asked for by 192 unique installs, ahead of webhook at 137.
Changes versus the previous issue
| figure | value | change |
|---|---|---|
| Workflows created | 80,197 | — |
| Workflows per day | 11,457 | +10% |
| Tool calls | 2,054,044 | +22% |
| Calls ending in error | 0.10% | -25% |
| Failed validation | 0.7% | -8% |
| Active installs | 151,238 | +6% |
| Week-1 retention | 45.2% | — |
| Workflows added to corpus (week) | 31,558 | +22% |
This week's answers
- 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%.
- 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.
- 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. 2.3% 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%.
- 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.
- 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.
- Which n8n nodes do AI agents search for?
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.
- 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.
Most searched nodes
data table (192) · Webhook (137) · http request (74) · Telegram (44) · google sheets (42) · schedule trigger (41) · anthropic (37) · whatsapp (34) · Slack (27) · gemini (27)
Webhook node response must be sent explicitly
When using the Webhook node with "Respond to Webhook" set to "Using Respond to Webhook Node", you must include a "Respond to Webhook" node in your workflow. Without it, the webhook caller will timeout waiting for a response. For simple acknowledgments, use "When Last Node Finishes" response mode instead.