API Health Monitor
Given an API endpoint + expected behavior, returns a structured health read: latency, status, expected vs actual response shape.
Example automations · as-is
DevOps & IT Ops
The old paradigm is out. The new paradigm is AI. AI automation is InTouch AI. The infra hygiene nobody schedules and everybody needs — and the one-off cron script that rots until it silently stops firing is dead. Describe the check. InTouch runs it, watches it, and when it breaks it reads the failure, tells you why, and heals what it can.
Backups, monitors, SSL/DNS watchers, log scanners, GitHub digests, postmortem drafts.
InTouch AI is a general-purpose automation platform. This page points it at DevOps / SRE / IT-ops work using example skills and jobs from the InTouch Hub. The same platform runs far beyond DevOps. These are starting points to adapt, not the ceiling.
The examples below are scheduled or trigger-driven jobs that run alongside your existing monitoring/CI stack. They don't replace Datadog or Prometheus. They fill the gaps where a custom script plus a cron job normally rots until it silently stops firing. The difference: the contract is now intelligent. You tell it what to do, when, what to do when it works, what to do when it doesn't, and who to notify — and the "doesn't work" clause stopped being a dumb rule. It's not "retry 3 times, email a log." It's an assessment: it reads the failure, smart-retries, refreshes the expired token, and surfaces the one sentence that matters. It broke. Here's why. I fixed it. No legacy cron job can say that.
These are examples, not turnkey products. Read the README and YAML before you run one. Most expect a Sheet, an HTTP endpoint, or an env var pointing at your infrastructure.
What you can automate
Things that quietly break
SSL certs expire Saturday morning. DNS gets changed by someone you don't remember authorizing. Disk fills up the day before quarter-end. Backups fail silently and you find out the day you need them. InTouch catches every one of these before they catch you.
Source-control noise
Your team pushed 47 commits this week, opened 12 PRs, merged 9. Weekly digest of activity beats opening GitHub. Release notes drafted from commit messages. Bad commit message? Catch it on push.
Log triage
Your prod logs have 3,200 ERROR lines yesterday. Half are duplicates, half are real. Scan, dedupe, surface the actually-new ones. Postmortem draft from an incident thread.
Cloud bills you don't watch
Your AWS bill jumped 40% last month and nobody noticed until the invoice. A daily Cost Explorer pull and a threshold alert catch the runaway instance while it's still cheap to kill.
Featured example skills
You don't start from a blank page. Find it in the Hub, install it, point it at your setup, run it — in your own language. These are starting points that already work.
Log Error Scanner
Paste a log slice; get a deduped summary of the actual unique errors, frequencies, and likely root causes. Better than `grep ERROR | sort | uniq`.
Error Postmortem Drafter
Given an incident timeline + Slack thread, drafts a postmortem skeleton: what happened, contributing factors, timeline, mitigations, action items.
GitHub Activity Digest
Summarize a repo's week: commits, PRs opened/merged, issues filed/closed, notable changes. Useful for weekly engineering reports.
Commit Message Improver
Take a vague commit message ("fix stuff", "WIP"); return a structured conventional-commit style message based on the actual diff.
Repo Release Notes
Draft a release notes section from commits between two tags. Categorize by type (features, fixes, breaking changes).
Featured example jobs
SSL Cert Expiration Alert
Daily check of a list of hosts; alert at 30/14/7 days before any cert expires. Sheet-backed list of hosts to watch.
DNS Change Detector
Periodic check of DNS records for a list of domains. Alert on any change (A, CNAME, MX, NS). Catches unauthorized changes early.
Website Uptime Check
Periodic HTTP HEAD against a list of URLs. Alert on non-200 or slow responses. Cheap external monitor.
Backup Verifier
Trust but verify: scheduled check that your backups (S3, local NAS) actually contain recent files, not just an empty success status from yesterday's cron.
Disk Space Alert
Per-host disk fill alerts. SSH out, run df, alert if any mount is over threshold. Headless servers especially.
Code Review on Change
When a PR is opened, post a Claude-drafted code review comment with suggested improvements. Augments human review, doesn't replace it.
S3 Cleanup
Scheduled cleanup of S3 prefixes older than N days. Prevents your dev/staging buckets from drifting to thousands of dollars/month.
AWS Cost Watcher
Daily AWS Cost Explorer pull. Alert if today's spend trajectory exceeds a threshold (catches stuck-on EC2 instances and S3 explosion early).
Database Snapshot Report
Daily summary of your DB: row counts per table, growth, key metrics. Spot anomalies before customers report them.
How it works
1. Install
InTouch AI runs as a single JAR on any host with Java 17. Or in a container, or on a Kubernetes pod. Self-hosted, laptop to enterprise — your creds and logs never leave your network. Two doors in: find-and-run in your own language, or jobs-as-code, MCP, RAG, agentic Monitors for the builders. Free Personal edition for evaluation.
2. Adapt
Each job reads from either a Sheet, an HTTP endpoint, or an env var. Read the README, set the placeholders, point at your infrastructure.
3. Schedule + alert
InTouch runs the job on cron-style schedules, logs every run, and alerts on your existing channels (email, Slack, Discord, Teams, PagerDuty webhook). When something breaks, it reads the failure, fixes what it can, and surfaces the one sentence that matters — not a stack trace. You didn't write it line-by-line, so you can't debug it line-by-line. It debugs it for you.
Where we don't pretend
- Not a Datadog/Prometheus replacement. InTouch is built for "watch X, alert Y, on a schedule." For high-volume metric scraping, real-time dashboards, and distributed tracing, use the right tool. We won't lie about that.
- Not a CI/CD replacement. Keep Jenkins, GitHub Actions, GitLab CI. InTouch is the layer that watches THOSE and tells you when a build fails and why.
- Not a SIEM. Log scanning here is weekly-hygiene grade, not security-operations-center grade. Know the difference before you lean on it.
Run it tonight
Personal edition is free. One JAR on your dev box or homelab server. Take one example, point it at your infrastructure, and watch it save you an hour this week. And you hold the determinism dial: run a check AI-native while it earns trust, then graduate it to fully deterministic — zero AI cost per run, identical every time, fully audited. Stop configuring. Start describing.