SafeGrd console
02 / Surfaces
A Fire Drill restores a backup somewhere disposable, checks what came back and records the result.
- :
Not backed up
New host token
The old token stopped working. Run this on to connect it again. The token is not shown again.
On a rebuilt host, install safegrd first:
The host keeps backing up to its key. A SafeGrd-managed key is fetched for it. For a customer-managed key, put the key file at ~/.safegrd/keys/daemon.key first. Then run safegrd claim to add the surfaces the console names for this host.
Loading your surfaces
Nothing protected yet
Choose where backups go and what to protect. A host you connect backs up a database on your server. SafeGrd backs up a managed one (Supabase, Neon) from its connection string.
| Surface | State | Last 30 days | Verified point ? | Restore time ? | Schedule | Details |
|---|---|---|---|---|---|---|
| No surface is in that state now. | ||||||
|
|
|
|
|
Paused
Change pending
|
||
03 / Hosts
No hosts yet
A host is a server where safegrd runs your backups. A surface SafeGrd backs up, or a WordPress site, needs none.
| Host | Runs | How | Last check-in | Version | Details |
|---|---|---|---|---|---|
| No surfaces yet |
No daemon. Backs up from cron or by hand.
|
Silent
Not backing up
If this host is gone, restore from a new one, then retire it.
|
Update available
|
|
01 / Overview
Nothing to show yet
Each surface's last 30 days show here once it is added.
| Surface | Last 30 days | Verified point |
|---|---|---|
Alerts
Loading
A failed backup or drill, a host that stops checking in, and storage running low are listed here.
To get alerts where you work, add a webhook or turn on email under .
Loading this surface's snapshots and drills
No backup in the last 48 hours.
safegrd wal base) to start an unbroken window after it.
To back up now, run on the host.
Running safegrd daemon there adds a Back up now button here.
| Taken | Size | Drill | Retention | Restore |
|---|---|---|---|---|
| Failed Verified Not drilled |
| Result | Restored | Tested | Certificate |
|---|---|---|---|
No restore has been asked for from the console. and press Restore on a backup.
No restore has been asked for from the console.
| Result | Restored | Into | Asked by | Record |
|---|---|---|---|---|
|
|
In your local time, inside the window.
That moment is outside the window.
On a host with this surface's config and key, into an empty database. It loads the newest base backup before that moment and replays the to it:
On a host with this surface's config and key, as the PostgreSQL user:
On a host with this surface's config and key. It writes a new database file, to within a second of that moment:
The window's end is when the last WAL file was pushed. PostgreSQL refuses it as a time when nothing was committed after it, so this command restores everything archived instead. Pick an earlier moment to restore to a time.
On a host that replaces this one, add when the archive is on hosted storage.
This surface restores to its backups only. Give it a wal: block in the host's config, and the resident daemon follows the database's WAL, so it restores to within a second of any moment in a window.
The database has to be in WAL mode, and the daemon needs write access to it and its directory. In your own bucket on every plan; on hosted storage from the plan up. Set up point-in-time recovery
This surface restores to its backups only. Give it a wal: block in the host's config, and the resident daemon follows the database's , so it restores to any second in a window.
The server needs its binary log on in ROW format, and the surface's account REPLICATION SLAVE and REPLICATION CLIENT (BINLOG MONITOR on MariaDB). In your own bucket on every plan; on hosted storage from the plan up. Set up point-in-time recovery
The server has to be a replica set (one member is enough), and the surface's user needs read on the local database. In your own bucket on every plan; on hosted storage from the plan up. Set up point-in-time recovery
This surface restores to its backups only. Archive its write-ahead log to restore to any second in a window. On the host, give the surface a wal: block in the config, and set these in postgresql.conf:
The surface id is the one in the host's config, not this name. Self-hosted PostgreSQL only. In your own bucket on every plan; on hosted storage from the plan up. Set up point-in-time recovery
Loading tables
No backup has recorded tables yet.
| Table | Rows | Newest row | Schema | Last drill |
|---|---|---|---|---|
|
RLS
|
Not drilled |
Not recorded. Update the CLI on the host.
| Column | Type | Constraints |
|---|---|---|
- Foreign keys to: (not loaded)
- Owns:
- No foreign keys or owned sequences.
safegrd restore --table How to restore one table
04 / Snapshots
No snapshots yet
Run safegrd backup to create your first encrypted snapshot.
| Snapshot ID | Surface | Contents | Size | Threat Shield | Retention | Created | Restore |
|---|---|---|---|---|---|---|---|
|
|
|||||||
| Host removed | Files Email Postgres MySQL MongoDB SQLite WordPress Laravel |
Failed
Anomalous
Clean
First backup
Verified
Incremental
Full copy
Outside project storage
|
Kept as last known good
|
||||
05 / Fire Drills
Each drill restores a snapshot and checks what came back. The result is recorded against the surface.
You are this organization's auditor. Each drill below restored a backup and checked what came back, and SafeGrd signed the record. The Evidence Pack collects them for a period.
Each surface's next drill: when, where, and how often the plan drills. Overdue first.
No surface has a backup to drill yet.
Loading
| Surface | Next drill | Runs on | Cadence | Next backup |
|---|---|---|---|---|
Recovery point and restore time per surface, measured from its snapshots and drills. What RPO and RTO mean. How each figure is measured.
No backups yet
Recovery point shows after the first backup, restore time after the first drill.
One cell a day, Monday first, UTC. Filled: a passing drill. Alarm: a failing one.
No surface yet.
No Fire Drills recorded yet
Each drill the organization's hosts or SafeGrd run is listed here once it is recorded.
No Fire Drills yet
Run safegrd verify --snapshot … to restore a snapshot in a sandbox and prove it comes back.
| Status | Surface | Restored | Tested | Certificate |
|---|---|---|---|---|
| Public Ran on SafeGrd |
A badge is available once a surface is protected.
README badge and public page
The badge shows whether drills pass. Publish the surface's page and it links to the newest certificate.
Shows the newest passed drill and updates as new ones pass. Each one shown is published.
An owner or admin can publish this surface's page.
Project pages
One page for every surface in a project: its kind, last backup and newest drill. The page says how many surfaces the project has.
An owner or admin can change this.
07 / Access tokens
Tokens authenticate the CLI, GitHub Actions and Kubernetes sidecars. A token can enroll a host, read, and ask for a backup or drill. It cannot delete backups, remove surfaces or change storage.
Token created
Copy the token now. It is not shown again.
No access tokens
Shown once, at creation. Expires after the time you choose, up to a year.
| Token name | Key prefix | Scope | Created at | Expires | Actions |
|---|---|---|---|---|---|
| No expiry |
Backup and drill requests
Every Back up now and Drill now: who asked, how, and the answer.
No requests yet
Scheduled backups and drills are not listed here, only the ones someone asked for.
| When | Request | Surface | Asked by | Outcome |
|---|---|---|---|---|
06 / Projects
Projects group surfaces and set where their backups go. Unlimited on every plan.
No projects yet
A project keeps one environment's backups apart from another's, with its own storage and its own hosts.
| Project | Slug | Status | Locks | Storage | Surfaces | Actions |
|---|---|---|---|---|---|---|
| Archived Active |
Hosted storage
|
|
Project storage
SafeGrd-hosted storage, where the plan includes it, or your own S3-compatible bucket. Every snapshot is encrypted before upload, so storage only ever holds ciphertext. Choose another project with its Storage button above.
Backups keep going to hosted storage until a bucket is saved below.
S3 storage
Every backup of this project is also written to each bucket below, under its own key and lock. A bucket that is behind is caught up by the host.
| Name | Bucket | Lock | Key | Actions |
|---|---|---|---|---|
Organization storage
Hosted storage and the key access record cover every project in this organization.
Counts every project on hosted storage against what your plan includes.
| Surface | Project | Written per day | Held for | Holds now | Settles at |
|---|---|---|---|---|---|
|
Decommissioned. Writes nothing more.
|
Organization storage
Hosted storage and the key access record cover every project in this organization.
Key access record
Secret releases and refusals across the organization. Secrets are named by type, never shown.
Nothing released yet
No stored secrets have been released to any host in this organization.
| When | What | Host | Project | From | Outcome |
|---|---|---|---|---|---|
08 / Activity
Owners and admins see this organization's activity log.
Key access record
Every key and credential SafeGrd released, to which host and when, and every refusal. Reads of hosted storage are listed there and not in the activity log.
09 / Billing
Every paid plan protects unlimited surfaces. Plans differ in drill cadence, whether a drill loads a real database, and evidence export.
Every paid plan:
If you cancel:
Without a plan after the trial:
Payment
Renewals are charged to this card.
No card on file.
Invoices
No invoices yet.
| Date | Total | Status | Invoice |
|---|---|---|---|
Charges stop at the end of the current period. Backups continue and no snapshot is removed. No new surfaces afterwards. Resume any time before the period ends.
11 / Help
Support
Write to support@safegrd.dev, or use the contact form. Paste the diagnostics below into the message: support asks for them first.
What's new
Every change to SafeGrd, newest first, is on the changelog.
10 / Settings
Theme
Profile
Two-factor sign-in
Sign-in methods
Organization
Security policy
Key access record
Every key and credential SafeGrd released, to which host and when, and every refusal.
Activity log
Sign-ins and changes in this organization, with who made them, are under Activity.
Alerts
SafeGrd alerts you when a backup or drill fails or is blocked, a host goes silent, a surface is overdue or storage runs low, and again when each recovers.
PagerDuty:
Email:
Saved when you change it.
Owners and admins change where alerts go.
A failure or a warning opens an incident on the PagerDuty service, and its recovery resolves it. Where to find the URL in Slack, Discord or Teams, and the key in PagerDuty
Compliance platform
Each Fire Drill record is sent to your compliance platform as signed, dated evidence with a verify link. SafeGrd keeps the API token sealed.
:
Sending evidence to a platform comes with the plan, with the Evidence Pack.
Team members
Team members come with a paid plan
Free is for the owner alone. Paid plans add unlimited admins, members, viewers and auditors.
| Name | Role | Sign-in | Added | Actions | |
|---|---|---|---|---|---|
| Unverified - |
|
Delete organization
Deleting removes this organization's projects, members and settings, and signs you out.
Retire every host and cancel the subscription first. Backups already written stay in their storage.
Copy hosted ones out first with safegrd export. We keep the organization's records for
12 months so it can be restored, then erase them. To have them erased sooner, write to
support@safegrd.dev.
Platform overview
Global telemetry, multi-tenant directory, and early access waitlist across all organizations.
Stale: no pass for three intervals, or the last three failed. The operator channel is told when it starts and ends.
| Job | Every | Last run | Took | Last success | Next run | State |
|---|---|---|---|---|---|---|
This server process, what ops-watch has open, and what the last 24 hours recorded. Errors cover the last 10 minutes, which is what the process keeps.
Open in ops-watch
Nothing is open.
| Since | Severity | Rule | Subject | Detail |
|---|---|---|---|---|
Last 24 hours
The file store keeps no account or billing tables to count from.
| Signups | |
| Password links sent | |
| Refused sign-ins | |
| Billing webhooks |
Database
Backups and drills SafeGrd runs on its own machines: every open job, and every job that ended in the last 24 hours.
The queue is the runner_jobs table. The server sweeps it every
seconds, at start, and as soon as someone asks for a job.
| Ran on | Ended | Failed | Wait p50 / p95 | Boot p50 / p95 | Run p50 / p95 / max | |
|---|---|---|---|---|---|---|
| No job ended. | ||||||
| Queued | Organization | Surface | Job | State | Ran on | Machine | Detail |
|---|---|---|---|---|---|---|---|
| Step | Reached | Of signups | Lost at this step |
|---|---|---|---|
By source
| Source | |
|---|---|
| No sign-ups in this window. | |
Not finished yet
| Organization | Signed up | Got as far as | Source |
|---|---|---|---|
| Tenant organization | Owner | Plan / Tier | Surfaces | Hosted storage | Latest backup | Fire Drill health | Created | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
|
Not offered
|
|
|||||||||||||||||
|
|||||||||||||||||||
Periods the hourly sweep left for a person. A period marked sending was being charged when the server stopped, so whether the provider charged it is unknown. Check the provider, then answer: Charged records it as sent, Not charged sends it again on the next sweep. Abandoned and unsupported periods are invoiced by hand, then marked charged.
| Organization | Period | Plan | Overage | Provider | State | Last answer | |
|---|---|---|---|---|---|---|---|
| Work email | Name | Company | Surfaces / Notes | Joined at | Status |
|---|---|---|---|---|---|
| Table | Rows before | Rows now | Newest row |
|---|---|---|---|
No tables added, removed or changed.
The host that took this snapshot was removed. Restore it from any enrolled host of this organization with .
This is not the storage this surface's project uses. The backup is recorded, and a restore reads it from the location above.
To keep it longer, run on a host that writes to this bucket: (at most 13 months ahead).
File types ( types • files across directories)
| File extension | File count | Aggregate size |
|---|---|---|
Mailbox folders ( folders • messages)
| Mailbox folder | Message count | Folder size |
|---|---|---|
Tables ( tables)
| Schema | Table name | Row count | Size | Newest row | Uploaded |
|---|---|---|---|---|---|
| Row-level security |