System information
System information is a short, per-user list of things that need your attention. It is not a notification feed. A notification says something happened and is read once; a system information item describes the current state of your account or team and goes away by itself when the cause is fixed.
There are two kinds of item.
- Action items are derived from the current state and carry a link to fix it. Today: a git provider connection that Canopy can no longer use (an expired or revoked token, or a provider that refuses the access Canopy needs), and, for team owners and admins, a team clone identity that is about to expire or has expired. Reconnect the provider, or rotate the clone identity, and the item disappears. If the problem comes back the item comes back, even if you dismissed it before.
- Announcements are written by people: Canopy staff can address everyone, one team or one user; a team admin can address their own team.
How items are shown
Section titled “How items are shown”| Severity | Meaning |
|---|---|
info |
Good to know. |
warning |
Something will stop working if you do nothing. |
critical |
Something has stopped, or will very soon. |
An item is a banner or card unless it is blocking. Only a blocking action or a critical announcement can be blocking; clients show those as a modal.
You can dismiss an item you do not want to see again, or snooze it until a later time. A blocking action cannot be dismissed, only snoozed for at most 24 hours. Anything else can be snoozed for up to 30 days. An announcement can be published as not dismissible or not snoozable. A dismissed action item stays hidden until its cause is fixed and later returns.
The first action items
Section titled “The first action items”| Key | Shown to | When | Fix |
|---|---|---|---|
git_provider_link:<provider> |
the account owner | Canopy’s last use of your stored token was rejected by the provider (401), the stored token cannot be read, or a periodic check found the provider refusing the account (403) | Reconnect the provider under Settings, Connected apps |
clone_identity:<id> |
owners and admins of the identity’s team | the identity is within 14 days of expiry (or past its own reminder date), or has expired | Rotate or register a new identity under Team, Git access |
Canopy records a rejected or unreadable token when a Source read or the building of your provider client hits it, and re-verifies every stored provider connection about once a day in small batches, so the list does not depend on you using the connection. Nothing is checked when you load the list. Reconnecting the provider clears the signal at once.
None of the first action items is blocking.
Read your list
Section titled “Read your list”GET /me/system-information{ "items": [ { "id": "4f0a3c1e-7a5b-5c0d-9d3a-0c1b2a3d4e5f", "key": "git_provider_link:gitlab", "kind": "action", "severity": "warning", "title": "Reconnect your GitLab account", "body": "Canopy can no longer use your GitLab connection, so your GitLab repositories cannot be read until you reconnect.", "cta": { "label": "Reconnect GitLab", "href": "/settings/connected-apps", "action": "reconnect_git_provider" }, "blocking": false, "dismissible": true, "snoozable": true, "starts_at": "2026-10-19T08:00:00.000Z", "ends_at": null, "state": { "seen_at": null, "dismissed_at": null, "snoozed_until": null, "completed_at": null } } ]}Only items that are active, not dismissed and not currently snoozed are returned: blocking first, then by severity,
actions before announcements, at most 50. The response is sent with Cache-Control: private, no-store. An action
item’s id is stable: the same cause gives the same id, so state you recorded stays attached. cta.href is an https
URL or an app path starting with /; cta.action names a client-side action such as reconnect_git_provider.
The current derived-action checks return at most 10 unhealthy provider connections and 20 expiring clone identities
per user, before the overall feed limit is applied.
POST /me/system-information/{id}/seenPOST /me/system-information/{id}/dismissPOST /me/system-information/{id}/snooze { "until": "2026-10-20T08:00:00Z" }Each answers 200 with { "id": "...", "state": { ... } }. 404 for an item that is not active for you. 409 not_dismissible for a blocking action or an announcement published as not dismissible (not_snoozable for the
same on snooze). 422 invalid_until when until is in the past, malformed or beyond the window above. State routes are
limited to 120 requests per minute.
Publish a team announcement
Section titled “Publish a team announcement”Team admins (team:manage, admin and above) publish to their own team only:
POST /teams/{team_id}/system-informationGET /teams/{team_id}/system-information ?status=active|ended|all&limit&offsetPATCH /teams/{team_id}/system-information/{id} { "ends_at": "..." } (optional; ends it now by default)Ending an announcement sets its end to now (or to the given ends_at) so it leaves every list.
| Field | Rule |
|---|---|
title |
required, plain text, at most 140 characters |
body |
optional, plain text, at most 2000 characters |
severity |
info (default), warning or critical |
blocking |
default false; only with severity: critical |
dismissible, snoozable |
default true |
starts_at, ends_at |
ISO dates; starts_at defaults to now, ends_at must be in the future and after starts_at, and none means “until ended” |
cta |
{ label, href }, label at most 60 characters, href an https URL or an app path starting with / |
HTML-like tags, control characters and link schemes other than HTTPS or an app path are refused with 422 invalid_announcement, listing each field at fault. Publishing and ending are limited to 30 requests per minute per
user, and every publish and end is written to the audit log (system_information.published,
system_information.ended). Staff announcements use protected administrative workflows documented in the
internal operations guide.
From the CLI
Section titled “From the CLI”canopy system-info [list] [--json]canopy system-info dismiss <item_id>canopy system-info snooze <item_id> [--until <iso8601|2h|3d>]canopy system-info publish --team <team_id> --title <text> [--body <text>] [--severity info|warning|critical] \ [--blocking] [--not-dismissible] [--not-snoozable] [--starts <when>] [--ends <when>] [--cta-label <text> --cta-href <url|/path>]canopy system-info announcements --team <team_id> [--status active|ended|all]canopy system-info end <announcement_id> --team <team_id> [--ends <when>]--until, --starts and --ends take an ISO 8601 time or a duration such as 90m, 2h or 3d; a snooze without
--until lasts 24 hours. canopy whoami and canopy login print one line on standard error, for example 2 items need your attention (1 blocking): run "canopy system-info", when something is open. They stay silent with --json or
--quiet, when output is not a terminal, and when the API cannot be reached. No other command prints it.
From the SDK and MCP
Section titled “From the SDK and MCP”const items = await sdk.systemInformation.list();await sdk.systemInformation.snooze(items[0].id, new Date(Date.now() + 3600_000));await sdk.systemInformation.team.create(teamId, { title: "Deploy freeze", severity: "warning" });An agent can read the list with the read-only MCP tool host_list_system_information (at most 20 items, long bodies
cut at 280 characters). Dismissing, snoozing and publishing are not available over MCP.