Get clean output the first time.

1. Sign up at /sign-up. The free tier (5 runbooks/day, no card required) is enough to evaluate the output on a real incident. Upgrade to Pro any time for the rest of the doc types and features.
2. On the app page, pick a doc type (runbook, post-mortem, SOP, or architecture), paste your raw input, and click Generate. Output streams back formatted in real time.
3. Review the output, refine any section that needs more detail, and export in whichever format your team uses.

The prompts are built for messy, unstructured artifacts — Slack transcripts, pager timestamps, terminal scrollback, scratch notes. You don't need to clean them up first. The four things that move the needle:

• Include the exact commands you ran, flags and arguments included. The prompts preserve them verbatim — no paraphrasing your known-working command into something close.
• Include timestamps (UTC is best). The post-mortem prompt builds a timeline table from them; the runbook prompt uses them to sequence diagnosis and resolution steps.
• Name specific systems and services. If you say "the api", you'll get "the api". If you say "api-prod deployment in the billing cluster", you'll get that.
• Mention what you don't know, and it'll show up as [NEEDS INFO: ...] in the output rather than being invented. Runbooks with fabricated facts cause outages.

Each type has a dedicated prompt tuned for its output shape.

If your team already has a preferred voice or structure (numbering conventions, heading style, section ordering), upload an existing doc once and reuse it on every future generation.

1. Go to Templates. Name the template, paste or upload a .md/.txt file.
2. On the generator, switch the Style radio from "None" to "Style template" and pick the one you added.
3. The template is passed to the model as a clearly-delimited reference block, tagged as a style example rather than as instructions. The doc-type prompt rules still apply.

After a generation completes, hover any ## heading in the rendered output. A Refine ↗ button appears. Click it, describe what you want to add or change, and submit. That one section regenerates with the new context; everything else stays untouched.

Refinements are a new row in your generation history, linked to the parent via parent_generation_id. So you can trace how a doc evolved.

Every generation is stored on your account. Visit /app/history to search, filter by doc type, and re-download any past generation in any format.

Deletion. Each history row has a Delete button. Deleting cascades — input, output, and all derived usage stats are gone. Deleting your whole account (via Clerk in the user menu) cascades to every row you created.

PII awareness. Inputs often contain secrets, internal URLs, and sensitive system names. The prompts redact secret-shaped tokens from the output, but the input we store is raw. A scheduled auto-purge after N days is on the roadmap; for now, delete manually what you don't want retained.