AvailableRuns in your browser
Terraform Plan Analyzer
Find destructive and high-risk changes in a Terraform plan before you apply it.
Terraform · OpenTofu
Workbench
The workbench
Tools we built for our own operations work. The ones marked available run entirely in your browser — pasted plans, configs, and command output never leave your machine, which is also why they work without an account. The ones marked as not yet built are listed so the roadmap is visible, not because they do anything.
Available
Working tools
Each of these does deterministic analysis — rules and parsing, no model inference. That makes them fast, private, and predictable.
AvailableRuns in your browser
Docker Production Auditor
Check a Dockerfile against production-readiness rules.
Docker · OCI · Containers
AvailableRuns in your browser
OpenStack Health Analyzer
Turn service and agent listings into a ranked view of what is actually broken.
OpenStack · Nova · Neutron
AvailableRuns in your browser
Alert Rule Generator
Generate Prometheus alerting rules that will not page you for nothing.
Prometheus · Alertmanager · Grafana
AvailableRuns in your browser
Prompt Workbench
A searchable library of engineering prompts, kept inside the application.
LLM · Incident Response · Kubernetes
Not yet built
What we have not shipped
These need a server-side inference endpoint we have not stood up. Nothing behind them works today and we are not going to pretend otherwise.
Not built
Incident Analyzer
The intended workflow: describe the symptom, paste the diagnostic output you already have, and get back ranked hypotheses with the specific next command that would confirm or eliminate each one. This requires model inference on a server we have not yet stood up.
Blocked on
- — Server-side inference endpoint with request authentication
- — Rate limiting and abuse controls
- — A redaction pass so pasted logs never persist credentials
- — An evaluation set so hypothesis quality is measured rather than assumed
Not built
Infrastructure Review
Intended to review infrastructure definitions the way a senior engineer would in a pull request — reasoning about what the configuration will do under failure, not just whether it parses. Deterministic checks for some of this already ship in the Docker auditor; the general case needs model inference.
Blocked on
- — Server-side inference endpoint with request authentication
- — A rule corpus for Kubernetes, Compose, and Ansible
- — Benchmarks against known-bad configurations
Not built
Runbook Assistant
Intended to turn an incident write-up into a structured runbook draft following our runbook template. Drafts would always enter the editorial workflow at draft status — never published automatically, and never presented as verified.
Blocked on
- — Server-side inference endpoint with request authentication
- — Integration with the editorial workflow so drafts cannot bypass review
Why these run locally
Infrastructure output is sensitive, so we designed around not receiving it
A Terraform plan names your resources. A Dockerfile can contain a credential someone forgot to remove. Service listings expose your host naming and topology. The safest way to handle that data is not to hold it, so the deterministic analyzers ship as code that runs in your browser.
There is no upload, no request to an API, and no logging of what you paste. You can verify this: open your browser network tab and watch nothing happen when you press Analyze. The planned tools will need a server because model inference has to run somewhere — and when they ship, they ship with redaction and explicit retention terms.
Newsletter
Inside The AI Stack Brief
A practical weekly briefing on AI engineering, infrastructure, production operations, and the technologies powering the AI stack.
One email a week. No sponsorship placements inside the technical sections. Unsubscribe in one click.