A cloud workspace's git acts as its solo owner, else the latest sender within the hour, else Orca's
bot; the control plane decides (#817's gitActor) and the desktop only says it, quietly:
- Source control, under Commit/Push: "Pushes and PRs go out as @login" / "as Orca bot" / "as Orca
bot (no one has sent in the last hour)", with the approvals/merges caveat in its tooltip. The
only place with an action: Connect GitHub, when it's the bot and the viewer isn't connected.
- The people chip: the same words as a second line in the acting member's avatar tooltip (the
first avatar's when it's the bot). No new header chip.
- Read with feature `git-actor` from GET /v1/team/workspaces/:id (owner and members alike), one
read per workspace shared by both places; re-read on focus, every 2 min, and when a shared
workspace's hour lapses (`until`). Absent field (not GitHub, identities off, older control
plane): nothing shown.
Archive and delete stop the VM but nothing it started elsewhere (a preview
database branch keeps billing). The control plane now runs scripts.archive
on the still-running VM: bash with set -e, as the user, with the session's
env and secrets, 2 min for all checkouts, output per checkout and in
~/.orca-sandbox/archive-hook/archive.log.
- POST/GET /v1/sandboxes/:id/archive-hook; migration 0062 remembers whether
the repo has a hook (learned at each sleep) so an asleep archive/delete can
offer a wake for it, and never wakes one without a hook.
- Desktop archive runs it before the Sleep; a failure shows the output with
Archive anyway / Keep it. Asleep: the Undo toast says it wakes first and
offers Skip hook.
- Asleep/unreachable delete runs it through the control plane; a failure
refuses with the usual Delete anyway waiver.
- Unattended archives (public API, team list) run it on a running VM and
archive even if it fails (logged).
- Docs: orca.yaml scripts.archive for Orca Cloud workspaces.
The wizard's applyCloudGitHubAccessToSandbox stub now calls #771's live refresh. Its IPC also
resolves the sandbox by the create's server environment (owned runtime record) or a GitHub-only
create's sandbox id (validated shape; the control plane's owner check is the guard), since the
create has no sidebar workspace yet.
MEMGUARD. A build or test that eats a cloud VM's memory now costs that process, not the workspace.
- orca-memguard (new VM entry point, root via sudo from orca-sandbox-start): ranks processes for the
kernel OOM killer (Orca runtime -900, terminal shells and agent CLIs -500, everything else 500),
stops the largest unprotected process when MemAvailable drops under 8% and records it, adds 2 GiB
swap files on demand (up to 1.5x RAM) and drops them after 10 quiet minutes, and remounts /dev/shm
at 50% of RAM (the provider mounts 64 MB). Reaches old snapshots through the entry-point refresh.
- orca-sandbox-drain releases on-demand swap before a stop so it doesn't ride the snapshot.
- Status read: the VM metrics probe reads the guard's state; status gains an optional memory
{pressure, guardKills, lastGuardVictim, ...} (metrics keeps its strict shape for older desktops).
- Desktop: guard kills count in the OOM tracker and show 'Orca stopped <process> to keep the
workspace responsive' with a Machine size action; the sidebar row gets a quiet memory mark under
pressure (amber when critical).
A per-person default Source 'in-workspace' for Claude and Codex. Orca stores and injects nothing
for that agent: no account placeholder, no API key, and so no egress route on the vendor's
hosts. The agent starts as it is and shows its own sign-in in the pane; its token stays in the
workspace's home directory and survives sleep.
- sandbox-control: migration 0062 widens agent_sources.source (table rebuilt, rows kept);
PUT /v1/agent-sources accepts 'in-workspace' as a default only (400 with workspaceId);
agentModesFor maps it to 'omit'; GET adds inWorkspace per agent while defaults/effective
keep their two values for older desktops.
- Desktop: coverage kind 'inWorkspace' (no hold, no per-host sign-in gate on Orca Cloud);
'Set up once' strip gets a quiet ⋯ choice 'Sign in inside this workspace instead' that
saves the choice and starts every launch held on it; Settings › Cloud agent sign-ins offers
it on a not-set-up row and shows 'Signs in inside each workspace' with a way back.
- Tests: Worker (env, open/limited policy, live switch, migration) and renderer (model, hold,
spawn gate, strip, settings, build row); fixtures for the strip menu and the settings row.
A create or new environment with nothing pinned used to start in pdx1 whatever the caller's location,
and every frame of a remote session paid that round trip. Now the sandbox-control Worker ranks the
served regions (the default and its fallbacks with an active golden) by distance from Cloudflare's
request geolocation and starts in the nearest.
- Explicit regions (request, API project) and an environment's build region stay authoritative.
- Environments record their region (migration 0061): builds, golden refreshes, staleness and the
security floor use it, so snapshots stay bootable and workspaces from them stay together. Legacy
environments (NULL) live in the default region as before.
- Automatic environments live where the create that made them ran.
- The capacity fallback tries the next nearest served region.
- The warm pool keeps VMs in every served region.
- VERCEL_NEAREST_REGION=off restores default-region-only.
- Desktop: the workspace card's Details shows the region (e.g. Portland (pdx1)).
Neil installed the Orca GitHub App on stablyai, and "Orca Cloud can't push to stablyai/orca yet"
with Check again never cleared: a running VM only got the App's repo token at boot (policy route +
the git placeholder header), so Check again re-ran a check that could never pass.
- sandbox-control: POST /v1/sandboxes/:id/git/access (owner) mints the repo token, swaps it into
the live egress policy, sets the VM git's placeholder header, and records the token expiry.
Answers applied, or not_installed / not_running / not_github. Idempotent.
- Desktop: Check again on the push and can't-reach lines, the poll after Install, and focus
rechecks apply the live refresh first, then re-run the check. If the App reaches the repo and
the check still fails after that, the line says "Restart the workspace to use the new access"
with Restart: never an endless Check again.
- No installation cache to go stale on our side: tokenForRepo caches minted tokens only.
Ask for Orca GitHub App access while the sandbox starts, not in a toast afterwards.
Server (sandbox-control):
- GET /v1/github/repo-access answers `push`: ok | not_installed | repo_not_selected |
not_connected, with `configureUrl` (the installation's settings page) for repo_not_selected.
A public repo that can't push now gets an install URL too.
- GitHub installation lookups (repo and new owner lookup) cached in D1 (migration 0061):
found 60 s, 404 15 s, dropped by installation / installation_repositories webhooks and the
setup return. The owner lookup never holds the answer (3 s cap, errors read as not_installed).
Desktop:
- "GitHub access" row after Start sandbox, checked when the create starts (parallel, so no
latency when access is fine). Missing: Install GitHub App / Add repository on GitHub /
Copy request for an owner, plus Skip for now ("You can pull; pushes need access later").
Polls (2 s for 3 min, then 10 s; on focus) and auto-advances on grant.
- The create holds before Open workspace only while the step asks and the sheet is on screen.
- Skip remembered per repo on this device; later creates show it skipped (Set up) without
holding; forgotten once the repo has access.
- A grant that lands after the sandbox started calls applyCloudGitHubAccessToSandbox, a stub
until gh-access's live refresh lands.
Neil: an Orca Cloud part inside Settings › Agent accounts was confusing. The Claude and ChatGPT
sign-ins for cloud workspaces (set up, paste a token, renew, remove) are now their own page in the
Orca Cloud group, right after Account & billing. Agent accounts keeps This computer and Hosts and
one row, "Orca Cloud sign-ins live in Orca Cloud settings", that opens the new page.
Every deep link moves with it: Renew in Settings, the sign-in sheet's footer, the host panel, and
old links to pane 'accounts' at 'orca-cloud' / 'accounts-orca-cloud' (resolveCloudAgentSignInsTarget).
Copy that sent people to "Settings → Agent accounts" for Orca Cloud now says "Settings → Cloud agent
sign-ins" (new i18n keys so stale translations don't point at the old place), including
sandbox-control's expired-Claude and RS-4 Codex messages. Search: the Orca Cloud keywords
(subscription, chatgpt, setup-token, renew) move to the new page. Behavior is unchanged.
The sidebar's Team list (PARITY-FOLLOW #652) is gone; the team view moves to the web console.
Following stays (now SidebarFollowingPanel/SidebarFollowingSection, the followed-workspaces
model). Its roster read, teammate activity model, cap, tests, fixtures and copy are deleted;
teammate profiles and the top-bar people are untouched.
Console links: n68 rejected tickets minted on console-staging.orca.dev because the allowlist was
only ORCA_CLOUD_CONSOLE_URL (app-staging) and the auth origin, so Team opened the bare console.
Auth now advertises `capabilities.consoleOrigin`; main accepts a ticket on that exact origin too
(https only, an origin and nothing more; the existing exact-match and no-userinfo guards stay), and
uses it for the no-ticket fallback. When a signed-in desktop's ticket fails or is rejected, the page
still opens but a toast says the console sign-in failed, instead of a silent sign-in page.
"Continue in browser" sends the plan and billing period picked in the plans dialog with the
console ticket ({page:'billing', plan, interval}) so the web billing page can preselect them. Main
adds the fields only when auth advertises the `console.billing_hint` capability: the ticket route
takes a strict body, and an older auth would refuse them and lose the signed-in landing. The
renderer's hint is re-validated in main (plan id pattern, month|year, billing page only).
The build sheet's no-plan blocker, after Compare plans, no longer says "Finish in your browser":
it says to choose a plan and check again, with See plans to reopen the dialog. Update payment
method keeps going straight to the console.
"Plan & billing", "Upgrade to Orca Cloud", "Choose plan" and every "Compare plans" now open an
in-app plans dialog first: the team's plan and status (trial hours left, active with hours and
seats, payment failed, or no plan with the trial billing offers), then billing's plan cards (Team
with a monthly/annual toggle, Enterprise with contact sales). "Continue in browser" opens the
console's billing page through the existing sign-in ticket. Members see the same plan info with
"Ask a team admin to upgrade" (Plan & billing is back in their account menu, read-only). Read-only:
no checkout, usage history or admin settings return to the desktop.
Everything shown is billing's: names, Stripe prices, seat pricing, feature bullets and trial terms
from GET /v1/plans, now also each plan's tagline, feature-list lead and button label (new optional
catalog `marketing`, unversioned copy, plus `omitFeatures` so Enterprise stops advertising egress
controls). Main's lenient parse keeps a newer or older service safe; without the catalog the
entitlement's offers stand in, and with neither the dialog says plans didn't load and still offers
the console. While billing loads the title is neutral, so a paying team never sees an upsell flash.
Settings > Account's billing row opens the dialog ("View plans"); "Update payment method" still
goes straight to the console.
The desktop's Routines settings are behind CLOUD_ROUTINES_UI_ENABLED=false, so the CLI
(orca cloud routine run) is where people read a run's result. A run that failed with Claude's
own "Not logged in · Please run /login" now says Claude isn't set up for Orca Cloud and where to
set it up, and doesn't repeat the raw summary. The detection moves to src/shared so both surfaces
use one rule.
Owner decision: paid and admin settings do not belong in the open-source desktop app.
Deletes the Enterprise pane (SSO, domains, SCIM, MFA policy, audit log), the Plan & billing
section and plans dialog (checkout, usage history, plan-feature-copy), and Team management
(members, invites, roles, leave, delete, create), with their hooks, fixtures, tests, i18n keys
and the main/preload/IPC only they used.
Each removed section is now one "Manage in browser" row. Main asks auth for a single-use
console code (POST /v1/desktop/console/ticket {page, orgId?} -> {code}) and opens
<consoleOrigin>/console/<section>?code=... with shell.openExternal; on any failure it opens the
page without a code. Endpoint, field names and origin live in cloud-console-config.ts
(ORCA_CLOUD_CONSOLE_URL overrides the origin, default is the auth origin).
Kept: sign-in, account identity and own MFA row, the team MFA gate/banners, the team switcher,
the roster (presence, names), plan status badge and payment notice, in-context billing
blockers (now "Compare plans" opens the console and re-checks on focus), cloud setup.
- Make owner confirmation lists what happens (sleeps, share links stop, the
disk goes with it, shared mode for good), stays open while the workspace
goes to sleep, and shows a refusal in place.
- Whoever made the change is told whether the previous owner will see a notice.
- The previous owner gets a notice (until dismissed) advising them to rotate
personal secrets; the new owner is told once that it runs like a shared
workspace.
- "Owner changed" chip in the top bar and a line in the sidebar hover card.
- Audit log words Orca's "previous owner told" row, and a row cut short shows
its whole sentence on hover (AUDIT-TRUNC-HOVER).
* feat(cloud): EP-10 delete a team whose plan is set to end: the dialog says it ends now with no refund; a failure after the plan ended says so and offers to try again
* fix(cloud): EP-10 delete dialog copy (design-squad)
* chore(cloud): EP-10 sync protected-copy keys for the delete dialog's new lines
* fix(cloud): EP-8 Delete of a running sandbox this computer can't reach (revoked pairing) goes through Orca Cloud with Force delete, never waits silently
* fix(cloud): EP-8 refusal copy says sandbox and this computer (design-squad)
Stripe's portal on API 2025-09-30.clover cancels by setting cancel_at, not cancel_at_period_end,
so billing never saw the cancellation: no 'Plan ends', no grant end. The mirror now stores
cancel_at (migration 0009); cancel_at_period_end is the 'cancelling' flag for either form; grants
end at cancel_at; entitlements carry cancelAt; the desktop's 'Plan ends' uses it.
Test: two teams of one owner (separate customers), one cancels via cancel_at: its own plan shows
ending on that date, the other is untouched; un-cancelling clears it.
* fix(skills): accept share links on the build's configured artifacts origin (EP-2)
A build signed in to staging publishes share-staging links, which the
install dialog rejected. The parser now also accepts the exact artifacts
origin this build resolves; production hosts are unchanged.
* fix(skills): declare shareLinkOrigins on SkillsApi
The Worker reads environment-build.sh's own lines from the stored build log at finalize and records
autoInstall { manager, outcome } in the build result (no image change). The build view carries it;
the environment page shows 'Dependencies installed (pnpm)' or that the install failed and workspaces
run their own setup.
* wip(cloud): PARITY-FOLLOW design fixtures (top-bar followers, sidebar Team and Following)
* feat(cloud): PARITY-FOLLOW phase 1: sidebar Team list and the top bar's people (creator and who is here now)
* test(cloud): PARITY-FOLLOW fixtures name the workspace a teammate is in
* fix(cloud): PARITY-FOLLOW fixture typecheck
* feat(cloud): PARITY-FOLLOW follow endpoint: workspace_follows (0055), PUT/DELETE /v1/team/workspaces/:id/follow, followers in workspace views
* fix(cloud): PARITY-FOLLOW lint: the hover action's wrapper owns its visibility; no Button restyles; no side effect in a state updater
* feat(cloud): PARITY-FOLLOW phase 2: Follow in the top bar, followers in the stack, Following section in the sidebar
* test(cloud): PARITY-FOLLOW no real-time wait in the live-read test
* feat(cloud): SSO-FOLLOWUPS members managed by the identity provider on the Team page; a clear message when single sign-on ends a sign-in
* fix(cloud): import the roster type where it is used
* feat(cloud): wording for single sign-on and directory sync events in the audit log
* fix(cloud): the missing-DNS-record audit line reads as an action (design-squad)
Conductor's cloud runs Amazon Linux; Orca's sandboxes Ubuntu. The Conductor importer now rewrites
the setup's own dnf/yum install lines to apt-get when every package has a clear Ubuntu name, and
leaves anything else as written. A script file the setup runs is only read (never changed or run)
to list its dnf lines. The setup-script review and the environment import show 'Uses dnf' with
what changed or the lines to check.
* feat(cloud): ENTERPRISE-SSO Settings › Enterprise single sign-on and directory sync, driven by server state
* feat(cloud): key icon on the single sign-on group title (TEAM-DIALOG-2)
* fix(cloud): Contact us in the SSO block is a plain button; Copy link is for sign-in surfaces
* test(cloud): the copied SSO setup link is minted like the opened one and never logged
* fix(cloud): design-squad review of the SSO blocks (sentence-case chips, directory sync icon, grace and last-change wording)
* fix(cloud): sentence-case success badge variant for the SSO chips; catalog matches the new copy
* fix(cloud): register the Enterprise IPC handlers with the team handlers (file line limit)