Paste an agent config, a cloud bill export you already own, or an observability inventory — get a reliability / smell / gaps report in your browser. Built by Arup Kumar Banerjee · Little Elm, Texas (Global SRE · LangGraph / AI agents · Observability · OpenShift). Also here: GLP-1 Support, a free, private organizer for people on or coming off GLP-1 medicines (educational only — not medical advice).
Why this exists · How to use it
Paste an agent config, system prompt or chat transcript. Engineering checks (retries, timeouts, tracing, secrets, injection sinks, tool auth, evals, LLM fallback) plus customer-safety red flags: made-up prices, missing handoffs, dosing advice, leaked notes, bots posing as people. Heuristic checker, not a security audit.
Open tool →Paste your AWS/GCP/Azure cost CSV or JSON export. Heuristic smells for egress, NAT, idle-ish lines, metrics creep. No cloud login. Heuristic checker, not a security audit.
Open tool →Paste a stack description or tick what you run. Gaps across OTel, Langfuse, Prometheus, Splunk + starter snippets. Heuristic checker, not a security audit.
Open tool →For people on or coming off Wegovy, Ozempic, Zepbound, Mounjaro and similar: symptom log, protein/fluids/fiber checklist, strength tracker, maintenance planner, insurance & refill checklist, questions for your prescriber. Not medical advice; data stays in your browser.
Open tool →A controlled local test: a scripted bot for a made-up dental office, 20 risky customer messages, and what the Agent Health Checker caught and missed before and after I fixed it.
Read the case study →What breaks in production is usually boring: an API key left in an agent’s config, a shell or SQL tool that runs whatever the model says (prompt injection is the first entry in the OWASP Top 10 for LLM apps), tool calls with no timeout, a service failing silently because nothing traces it, and a cloud or AI bill nobody can explain. That last one is growing: in a CloudZero survey of 260 finance leaders (June 2026), 87% said they need to tie AI spend to business outcomes within a year, but only 22% can today.
Checking for these is cheap. The catch is that the files you’d check are full of secrets. As an SRE, I wanted a first pass that never leaves the browser tab, so these tools have no backend. They’re heuristics, not an audit.
GLP-1 Support is for a different gap: most patients get medication instructions, but most aren’t referred to any other support. Read why it exists.
Agent config → Agent Health Checker. Cost export → Cloud Bill Smell Detector. Monitoring stack → Observability Gap Finder.
A config, a cost CSV/JSON you exported yourself, or a short stack description. Or tap “Load sample” to try a fictional one first.
Work down from high severity, then copy the Markdown report. Redact secrets before you share a screenshot.
Arup Kumar Banerjee · Little Elm, Texas — Global SRE background with LangGraph / AI agents, observability (Splunk, Langfuse, OpenTelemetry, Prometheus), OpenShift, cloud, and privacy-minded systems. These Labs tools are free public utilities for reputation and education — not consulting intake forms, not account scanners, and always-free public demos.
Scoring runs in your browser. Pasted configs and bill exports are not sent to a backend for analysis. No analytics or tracking scripts. Every tool here is free forever.
If a check looks wrong, tell me on GitHub. Please redact secrets before you post. A short snippet is enough — not a full config or bill export.