Self-hosted ops automation with eval loop
Suggested by runchr · today
Draft — a suggested stack, not yet verified by a member.
Build n8n workflows from node docs, then trace and evaluate the agent runs.
What they did
n8n-mcp gives the agent accurate node schemas and templates so it can assemble valid n8n workflows instead of guessing. The agent deploys those workflows into n8n, which runs the actual day-to-day ops automations. opik wraps the agent's runs in traces and evals, so failures and regressions surface as measurable scores rather than silent breakage.
How it went
A running set of n8n ops automations plus a trace-and-eval dashboard that shows which ones are degrading.
Worth knowing
n8n-mcp docs can lag the installed n8n version, and opik evals only catch what you thought to measure.
Wiring
Connect n8n-mcp to the agent first so it can read node docs and templates. Give the agent n8n API access to create and activate workflows. Then point the agent's LLM calls at opik for tracing and add eval runs on each workflow execution.
Try it yourself
I want to automate our weekly ops chores in n8n. Use n8n-mcp to build the workflows, deploy them, and set up opik tracing so I can see when one starts failing.
Tools in this stack
Run this stack
Install the stack into this project. It writes AGENTS.md and the manifest, merges the MCP servers into .mcp.json, and — with --yes — runs the tool installs.
npx runchr install runchr/self-hosted-ops-automation-with-observabilitycurl -s https://api.runchr.com/api/recipes/runchr/self-hosted-ops-automation-with-observability/manifestSystem prompt
You are an ops automation engineer who builds and maintains n8n workflows for a small company. You ground every node choice in n8n-mcp documentation and instrument each workflow with opik tracing.
The manifest is also at api.runchr.com/api/recipes/runchr/self-hosted-ops-automation-with-observability/manifest.