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.
- Last updated
- Oct 23, 2026
- Data window
- Oct 5–11, 2026
- Author
- Romuald Członkowski
- Mon 512,194
- Tue 613,540
- Wed 713,559
- Thu 813,874
- Fri 911,936
- Sat 107,490
- 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.
18% of n8n API connection checks fail, and 15% of those failures are a wrong address or unreachable network. That is the first thing to check when putting AI on an existing instance.
History
One row per weekly issue. The table grows with every update.
| week | workflows | calls | reads | writes | week-1 retention |
|---|---|---|---|---|---|
| 2026-W41 | 80,197 | 2,054,044 | 61% | 11% | 45% |
| 2026-W40 | 73,123 | 1,684,118 | 56% | 12% | n/a |
| 27 Sep 2026: telemetry source changed | |||||
| 2026-W37 | 118,918 | 2,502,508 | 60% | 14% | 57% |
| 2026-W36 | 98,043 | 1,784,850 | 60% | 15% | n/a |
| 2026-W35 | 98,122 | 1,870,624 | 57% | 15% | n/a |
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.
Methodology · Data and charts may be reused with attribution (CC BY 4.0).