For developers and QA

Write code, not docs.

Every merged PR becomes a docs update: API reference, changelog, and a plain version for everyone else.

Your week, before and after

Before

AI-assisted PRs land daily. The docs lag by weeks.

README, API reference and changelog all say different things.

"Quick question" DMs every time something ships.

You're the only one who knows how the billing service works.

After

Every merged change triggers a docs run.

API reference updated from the actual diff, not from memory.

Refactors, renames and dependency bumps skipped, with a one-line reason.

What you know is written down. The DMs stop.

Runs on merge.

The Sabí engine reads the diff, finds what changed for readers, and updates only those docs. No behaviour change, no update.

SabíDeveloper docs · Webhooks API
PATCH /v1/webhooks/:id
{
  "url": "https://example.com/hooks",
  "retry_policy": "exponential"New in 8.3
}
New field retry_policy. Retries 5 times with exponential backoff, max 1 hour.From change #2031

Customer steps, replayed in staging.

In beta: Sabí replays documented customer steps in your test environment and saves a screenshot per step.

SabíProduct check · in beta
The page saysAdd your URL and click Send test event.
Sabí tried it
Step works. Screenshot saved.

Read only. Sandboxed. Deleted.

Read-only access. A short-lived container with no network. Your full repository is never sent to the Sabí engine. Deleted after every run.

SabíSabí Secure Sandbox
Read onlySabí never writes to your repo.
No networkNothing inside the sandbox can call out.
Never the whole repoYour full repository is never sent to the Sabí engine.
Deleted after the runWhether it worked or not.

You stay in control.

Read-only access. Sabí never writes to your repo.

Every page carries a state: backed by the code, checked in the product, edited by a person, or failed.

Per area, choose auto-publish or human review.

Checked in the productBacked by a code changeNeeds a human lookDidn't work, held back
SabíProduct check · in beta
The page saysClick Replay failed events.
Sabí tried it
Held backNot published until someone looks.
A QA engineer laughing

A typical week

  1. Merge

    PR #2031 adds webhook retries.

  2. Run

    Sabí updates the API reference and changelog, and replays the setup steps in staging.

  3. Result

    Four steps pass. One button isn't switched on yet.

QA sees it before any customer does.

Make the case to your tech lead

The problem

You ship faster than anyone can document, so every "how does X work?" lands in your DMs.

What a pilot needs

Read access to your GitHub repos. For the beta product check: a staging URL and a test login.

What we measure

"How does X work?" questions to developers each week.

Send this page to your manager →

Questions

Which changes trigger an update?

Merged changes on GitHub. If behaviour didn't change (refactors, renames, dependency bumps), Sabí skips it and logs why.

Can it understand our codebase?

The Sabí engine is built for reading product code and writing docs from it. Try it on one of your public GitHub projects in the demo. Complex setups across many repos are next on our list to prove.

What happens to our code?

Each run analyses your code in an ephemeral, secured sandbox: read only, no network access, and deleted when the run ends. We keep the documentation and the short excerpts it cites, not your repository. And Sabí doesn't train AI models on your code.

What does the product check test against?

A test environment you choose, and only the test addresses you register. Never production. It's in beta, for customer-facing pages.

Can we get our content out?

Your docs are published into your own Notion, so they're already yours. A full workspace export is coming next.

See it on your own product.

Book a demoTry it on your code