> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sentivel.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Dependencies & cascade
> Show a third-party outage on your status page, in your own words.
Your service depends on other services. When Stripe, a cloud region, or an AI
provider has an incident, your customers feel it, and they ask you what is going
on. Sentivel watches those providers and, when one breaks, posts a plain-English
advisory on your status page automatically.
You map which of your components depend on which providers. That map drives
everything here. Sentivel never guesses a dependency; you define it.
## How cascade works
1. Sentivel polls a provider's status feed and detects an incident (or a part of
it going down).
2. It checks your dependency map for any of your components that rely on the
affected part.
3. For each match it posts an upstream advisory on your status page and flags the
component as affected.
4. When the provider recovers, the advisory clears on its own.
The advisory attributes the problem to the provider, in plain words:
> We're aware of an ongoing issue with AWS us-east-1, a service we rely on. Some
> functionality may be affected until it's resolved.
It never asserts that you are down, and never makes a promise it can't keep ("your
data is safe"). It states what is true: a provider you depend on has a problem.
## Link a dependency
1. Go to **Dependencies** in the dashboard.
2. Choose one of your components, then the provider it relies on.
3. Pick the specific parts you depend on (for example "Claude and Claude Code",
or a region like Cloudflare "Africa"). Leave the set empty to depend on the
whole provider.
An empty part set means the whole provider, so the component cascades only on a
filed provider incident. A part-scoped link also reacts to that specific part
going non-operational, even when the provider hasn't filed an incident yet.
A link re-evaluates as soon as you save it, so an issue that is already happening
shows up right away, not on the next poll.
## Auto-post or review first
Each cascade has two modes:
* **Auto-post (default).** The advisory goes up the moment a depended-on provider
has an incident. It's a factual attribution to the provider, so it needs no
sign-off.
* **Review first.** The cascade is flagged instead of posted, and Sentivel emails
your owners and admins to review it. You publish it with one click.
## Write your own advisory
Each dependency can carry its own advisory text. Leave it blank and Sentivel
writes the message for you, attributing the issue to the provider. Fill it in to
say it in your own words.
1. Open a dependency from the **Dependencies** page.
2. Type your wording in **Custom advisory message**.
3. Save. Up to 280 characters.
Your text replaces the automatic message while the upstream issue is open. The
resolution notice stays automatic. Name the provider, say what customers might
notice, and promise nothing you can't keep.
## Review what's live
The **Dependencies** page groups current upstream issues by provider. Each one has
a single reversible **Show** / **Hide** toggle, so you stay in control of what
customers see. Hiding an advisory is logged, and survives later polls of the same
incident.
## Suggestions
When a monitor's target is hosted on a provider in the catalog, Sentivel suggests
the dependency link. It never creates the link for you; you confirm it.