An organisation administrator can connect these under Hosting platforms, using an API token:
- Platform.sh (Upsun Fixed and Upsun): the form asks for the token only. Use a token from a dedicated user for Sentinel, not a person's own account. Give that user the Viewer organisation permission and access to only the projects you will map. Give it Contributor on the production environment if you want a backup taken before each deploy. Admin is not needed.
- Quant Cloud: the form asks for your Quant organisation machine name (as in the dashboard address) and an organisation API token that can read environments, change environment state and create backups. Add one connection per Quant organisation.
- Lagoon appears only if your deployment has switched on its preview setting.
Choose Connect for the platform, enter a label and the details, and select Connect. Sentinel checks the token with the platform before saving it. A token the platform rejects is not saved, and the message says so. The token is stored securely, write-only, and never shown again. When it works, the page lists the environments the token can see, by name and identifier only. Nothing is stored about them until a site maps one. Members see the list but not the controls.
To replace a token, rotate it through the API (POST /patching/hosting/connections/{id}/rotate); site mappings are kept. Remove revokes the token and is refused while a site uses the connection. If the platform later rejects the token, the connection shows an error and every site mapped to it leaves ready until you rotate the token.
If your platform is not offered, select My platform is not listed. It opens a form at Patching requests (site preselected when you start from a site's Patching tab, otherwise choose the site). The form says: "Integration with your platform is free of charge if you work with us to test it. Tell us about your account and we will propose the steps on this ticket."
It asks for the platform (Pantheon, Acquia Cloud, Shared hosting without an API or Another platform, with a platform name for Another platform), account type, subscription level, region, whether the platform has an API or CLI (and optionally a link to its API documentation), what triggers deployments (a git push, an action in the platform dashboard, the CI pipeline or not sure) and the number of sites. A known platform that is not supported yet shows a Known limitation note. Sending, ticket reuse, URL handling and who can send work as described under Connecting a Git host.
The site Patching tab
Open a site, then the Patching tab. What it shows depends on whether the site is enrolled (see Enrolling a site). Once it is enrolled it has the Readiness and Enrolled cards at the top, then the sections you switch between under Patching sections:
- Runs (the default): Run an update, Waiting on you when a pull request is open, and Past runs (see Run an update). While a job is queued or running, Runs shows a small live marker.
- Deployments: how updates reach production, with a Change in Configuration link, then each deployment (see Deploying a pull request).
- Deep scan: where the site's other checks and upkeep sit, beyond Composer runs (see Deep scan). It appears only once Sentinel has turned it on for your organisation.
- Activity: changes to the site's Patching set-up and who made them (see Patching jobs and activity).
- Configuration: the repository, branches, mode and schedule, package exclusions, the Runtime section (see Runtime updates), Deployment (who deploys), and Enrolment with Unenrol.
The section is part of the address (for example ?tab=patching&view=configuration), so you can bookmark or share it. Runs and Deployments say what to do when no repository is chosen yet, with a link to Choose the repository in Configuration.
Schedule. The configuration shows a Schedule heading with the organisation owner's time zone, one line for the security schedule and one for the maintenance schedule, Next maintenance runs and, when it applies, a warning that maintenance runs are further apart than the site's maintenance SLA. Administrators and site managers change the schedules in the Edit form (see Schedules); other members only read them.
Readiness and Enrolled cards. The two cards sit side by side at the top of the tab, Readiness first (stacked on a narrow screen), stay in place in every section, and start collapsed. Select a card's header to open or close it. The choice is not remembered, so each visit starts collapsed again, except where a card opens by itself as described below.
- Enrolled. The header shows the Copilot or Autopilot pill, Needs attention when it applies, the enrolment date and the profile name. Open it to read the full enrolment sentence and the reasons the enrolment needs attention. While the website is on Copilot, administrators and site managers also see Switch to Autopilot · merges and deploys for you under the header, which opens the Mode card in Configuration. They also see a link to unenrol the site from Configuration, where the Enrolment section has the Unenrol button. The card opens by itself when the enrolment needs attention.
- Readiness. The header shows the readiness badge, one dot for each checklist step (filled when it is done, hollow when it is still to do) and "N of M". Open the card to see each step by name in the checklist, with its links, the Deploy holds, the CI tests panel and the Platform not listed? link. The card opens by itself when the enrolment needs attention or a deploy hold is in effect.
Readiness checklist. A site is ready when all of these are done:
- An AI key is chosen and active (the site's own key, or the organisation default).
- A repository is chosen from a connected Git host. If the host stops accepting the connection, this shows as repository disconnected; the site keeps its repository and reconnecting restores it.
- The production hosting environment is chosen from a connected platform, or the site is set to "Sentinel merges, my pipeline deploys" or "I merge and deploy myself".
- Continuous integration (CI) has reported a result on a Sentinel pull request. Until then the badge shows a grey Waiting for first CI run.
- The repository's CI runs automated tests on every pull request and fails when they fail, or, on GitHub with the site set to "Sentinel merges, my pipeline deploys" and its Deploy workflow or check naming the workflow, the pipeline's deploy waits for the tests (see Automated tests in CI below).
- Your organisation has accepted the deployment terms.
The hosting step (step 3) stops Deploy only. It never stops Run update: when only hosting is missing, the tab says updates can run now. The CI step (step 4) stops nothing: it completes on its own the first time a CI check finishes on a pull request Sentinel opened, so run an update to get there. When it is the only step left, the tab says updates can run now. When another step still needs you, such as the terms, the badge shows that step instead. With "I merge and deploy myself" the step shows as done: "No hosting connection needed: you merge and deploy each pull request yourself". The pipeline choice needs no hosting connection either. A connection that is pending, in error or removed counts as missing, so a rejected key takes a site out of ready. The terms step has no link: you accept the deployment terms when you press Deploy on a pull request. Until the terms are published, that step stays open. Each unfinished step links to where you fix it.
