Let n8n be the deployment target for your agent
They used Claude to draft, edit, and ship n8n workflows through the API instead of building each node manually.
- What they did
- One commenter said they stopped assembling workflows in the n8n canvas and instead asked Claude to generate them from plain instructions. With n8n-mcp connected to Claude Desktop or the CLI, Claude could read and write to the n8n instance, create node setups from scratch, and revise existing workflows in smaller pieces. For larger rewrites, they moved to Claude Code, kept a CLAUDE.md file in the repo, worked on git branches, and let the agent handle longer refactors for later human review.
- How it went
- In their setup, n8n served as both the runtime and the place where workflows were deployed. That shifted the person’s job toward describing the behavior they wanted and checking the diffs instead of wiring nodes one by one. They did not report any concrete speed or scale improvements, and they warned that MCP can still be confidently wrong.
- Tips
- Changing only the named nodes was described as more dependable than replacing the entire workflow JSON or matching by name alone. Review every generated change before it goes live, especially anything that could affect production data or actions.
Try it yourself with n8n
I run n8n at [n8n URL]. Build a workflow for [task description], including any scripts or configuration files it needs, and deploy it through the n8n REST API with my credentials from [env var or secret name]. Create it in a disabled state first, then show me the workflow and wait for my approval before turning it on. If you need to edit an existing workflow, update only the specific nodes that changed. Do not send messages, make payments, delete anything, or submit anything live until I explicitly approve it. After that, run a test and tell me what the workflow does in plain language.
Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.
Get 3 agent uses like this every morning
With the day's top AI news. Free, and one click to unsubscribe.
Related agent uses
Set up an alert-diagnosis workflow for Kubernetes incidents
An n8n flow takes Alertmanager alerts, checks several MCP-connected systems, and writes a read-only diagnostic reply in Mattermost.
Auto-review GitHub pull requests with n8n
They used an n8n workflow to inspect new GitHub pull requests, generate a review with an LLM, and post feedback back to the PR.
Have the agent interview you before it writes code
A developer uses an agent to clear up vague requirements, handle small implementation choices, and keep them learning while the code is being built.
Review Linux kernel patches before they land
Google’s Sashiko scans Linux kernel submissions and flags bugs that human reviewers missed.
Turn Slack feedback into pull requests with an agent
Use an agent to sort Slack feedback, draft fixes as PRs, and help debug infrastructure with read-only access.
Review many PRs using a terminal agent
Meta's Muse Code checked 222 pull requests quickly and cheaply, but it also made a plan around a Google project that was not real.