vBackup MCP Tools

vBackup MCP Server exposes 68 tools with --allow-write and 41 tools in the default read-only mode. Every tool follows the EKS-style verb_noun naming convention (list_backup_policies, create_backup_server, delete_backup_destination).

Every tool carries an MCP ToolAnnotation:

Access	Meaning
read	Idempotent, no state change. Available in read-only mode. Clients can auto-approve.
write	Creates or modifies state. Requires --allow-write. Reversible (a schedule swap can be swapped back).
destructive	Deletes a destination, a policy, a backup server, a backup database or a restore point. Requires --allow-write. Clients should warn.

Data tools return structured Pydantic models — FastMCP emits outputSchema + structuredContent (JSON). Region is a fixed Literal["HCM-3", "HAN"]; every list_* output echoes back the region it was fetched from.

Catalogue and discovery

Tool	Access	Description
list_backends	read	Backends visible to the caller; id is the backendId every create needs
get_configuration	read	Platform limits a policy must respect — hourly intervals, retention ceilings, open hours
list_protected_servers	read	Instance IDs that already have a backup server (membership check only)

Backup destinations (the console's Backup Location)

Tool	Access	Description
list_backup_destinations	read	Where backups land, with quota, usage, soft delete and lock
get_backup_destination	read	One destination, read fresh — how every edit is verified
list_backup_destination_servers	read	vServer resources stored here; what a delete would destroy
list_backup_destination_databases	read	vDB resources stored here
list_backup_destination_tags	read	Tags on a destination; vng.* are platform-set and not editable
list_backup_destination_history	read	Config changes incl. failed attempts and the values used; account-wide when destination_id is omitted
list_backup_products	read	Products vBackup covers — the product string a create needs
list_backup_regions	read	Storage regions per product; a create takes region_id, not id
create_backup_destination	write	Create a location: product, region, quota, soft delete, lock
update_backup_destination_name	write	Rename; display name only, ids are unaffected
update_backup_destination_max_quota	write	Set the GB ceiling — runs FAIL once it is reached
update_backup_destination_soft_delete	write	Recycle bin; deleted backups stay billed for the retention
update_backup_destination_vault_lock	write	Retention lock — becomes permanent with changeDuration=0 or after the change window expires
delete_backup_destination	destructive	Delete the location and the backups stored in it

Backup policies

Tool	Access	Description
list_backup_policies	read	Policies with a flattened schedule.summary and how many servers use each
get_backup_policy	read	One policy, cadence by cadence — read before updating
create_backup_policy	write	Create a schedule (hourly/daily/weekly/monthly toggles + retention)
update_backup_policy	write	Replace a schedule wholesale; affects every server using it
update_default_backup_policy	write	Make one policy the default, demoting whichever holds it now
delete_backup_policy	destructive	Remove a policy; refused while servers are attached

Backup servers (protected instances)

Tool	Access	Description
list_backup_servers	read	Protected servers with policy, destination and per-disk flags; filter by server_id
get_backup_server	read	One backup server, read fresh (the list is cached)
list_backup_server_volumes	read	Which disks are included in runs, with size and usage
list_backup_server_points	read	Restore points a server holds, with the policy as it was at run time
get_backup_statistics	read	Account coverage and outcome counters — pass project_id or the ratio is meaningless
get_vserver_instance	read	The vServer machine behind a backup server (name, state, flavour, image)
get_backup_server_point_download_urls	read	Signed download links for a restore point — the links are credentials
create_backup_server	write	Protect instances with a chosen policy, destination and disk selection
update_backup_server_volumes	write	Include or exclude one disk from future runs
update_backup_server_policy	write	Attach a different schedule
update_backup_server_destination	write	Write future runs elsewhere; existing points stay put and stay billed
start_backup	write	Run a backup now ("Back now"), outside the schedule
enable_backup_server	write	Resume a paused schedule
disable_backup_server	write	Pause the schedule; restore points are kept and still billed
delete_backup_server_point	destructive	Delete ONE restore point
delete_backup_server	destructive	Delete the backup server and every restore point it holds

Backup databases (protected vDB instances)

The vDB half of vBackup. It shares policies, destinations and the history service with the backup-server family but nothing else — a database never appears in a backup-server listing, and each call answers 200 with an empty list rather than an error when asked about the wrong product.

Tool	Access	Description
list_backup_databases	read	Protected databases with engine, policy, destination and stored size
get_backup_database	read	One backup database, read fresh (the list is cached)
list_backup_database_points	read	Restore points a database holds, with stored and uncompressed sizes
list_protected_databases	read	vDB instance ids of one engine family that are already protected
list_databases	read	The vDB estate with per-instance eligibility and the reason when it is not
create_backup_database	write	Protect ONE vDB instance with a chosen policy and destination
update_backup_database_policy	write	Attach a different schedule
start_database_backup	write	Run a backup now, outside the schedule
enable_backup_database	write	Resume a paused schedule
disable_backup_database	write	Pause the schedule; restore points are kept and still billed
delete_backup_database_point	destructive	Delete ONE restore point
delete_backup_database	destructive	Delete the backup database and every restore point it holds

Four rules on the database family are easy to miss:

• databaseType is PostgresCluster or RedisCluster — a cluster spelling, not an engine name. An unrecognised value answers 200 with an empty list rather than an error.

• Only cluster deployments can be backed up. A single-node PostgreSQL is ineligible; list_databases says so per instance in ineligible_reason instead of silently omitting it.

• A backup location holds at most one backup database. Reusing an occupied one is refused with The backup destination already contains resources.

• A vault lock on the location blocks both deletes, answering Your resource is being managed by Vault. — unlike the "being processed" 409, retrying never helps until the retention expires.

History

Tool	Access	Description
list_backup_history	read	vServer backup runs newest-first; from_date — the default window is only 180 days
list_restore_history	read	vServer restores that have happened (this server cannot start one)
list_database_backup_history	read	vDB backup runs, with compressed vs uncompressed size
list_database_restore_history	read	vDB restores, and which instance the data was written into
list_server_migration_history	read	vServer migrations (Start / Complete / Rollback) — the status phase and the action that reached it are separate fields; read both

History is per-product and the two families never mix: a vDB run does not appear in list_backup_history. Configuration changes to a destination are a third, separate trail — list_backup_destination_history, which reads account-wide when destination_id is omitted.

vServer projection

The API publishes a second, vServer-flavoured copy of the backup-server endpoints. It is not a duplicate: it reports the captured instance's image and a restore point's per-disk detail, which the generic family does not.

Tool	Access	Description
list_vserver_backup_servers	read	Projection listing; project_id is required
get_vserver_backup_server	read	One backup server, projection shape (IAM-gated)
list_vserver_backup_server_points	read	Restore points with server_info — the image behind each point
get_vserver_backup_server_point	read	One restore point, projection shape (IAM-gated)
list_vserver_backup_volume_points	read	What is inside a restore point: which disk, which was bootable
get_vserver_backup_volume_point	read	One disk slice of a restore point (IAM-gated)
create_vserver_backup_servers	write	Shortcut create using the platform's default policy and destination
list_volume_usage	read	Size and used space of vServer volumes — a read expressed as a POST

Metrics (vMonitor)

Tool	Access	Description
get_backup_metrics	read	Product-wide trend: backup servers, servers, storage used — both regions in one call
get_backup_destination_metrics	read	Per-location trend: usage and success/failure counts; omit the id to chart every location

Both post fixed queries to the Backup Center's vMonitor API. vMonitor publishes exactly six vBackup metrics, so the payloads ship built in and the caller chooses only the time window — there is no free-form metric query, because an unknown metric name returns an empty 200 rather than an error.

Diagnostics and guidance

Tool	Access	Description
get_access_token	read	Current IAM access token, region and endpoint
get_feature_guide	read	Step-by-step guidance for a composite vBackup capability

IAM-gated tools answer 403 IAM_PERMISSION_DENIED for a caller whose policy does not grant them, while their sibling list tools work. A 403 there means "not allowed", never "does not exist" — each docstring names the list tool to fall back to.

Prompts

Nine guided flows are available as MCP prompts (Vietnamese) — portable across any MCP client, and always available (no --allow-write needed). Each prompt is also served through the get_feature_guide tool (same text, one source of truth): prompts must be loaded by the user, while agents call the tool on their own.

Prompt	Covers
vbackup_getting_started	The object model, regions and backends, and how vBackup differs from vServer snapshots
vbackup_protect_server	The full chain from discovery to creating a backup server
vbackup_protect_database	The vDB chain: eligibility, choosing a location and policy, and the delete constraints
vbackup_manage_policy	Schedules, retention, and the platform limits to validate against
vbackup_check_backups	Did last night's backup run, and why did it fail
vbackup_inspect_restore_point	What a restore point actually holds — and why a restore must be started in the console
vbackup_manage_backup_server	Coverage statistics, immediate backups, moving a destination, downloading and deleting points
vbackup_manage_destination	Creating, editing and deleting a backup location — and which settings become irreversible
vbackup_reduce_backup_cost	Finding backups that cost money for nothing

Known limits

• A restore cannot be started through this API. The gateway publishes list_restore_history and no endpoint to trigger a restore; restores are performed in the console.

• list_volume_usage 404s on a deleted server. The endpoint answers 404 "Not found volumeId <id>" for the whole request when any requested volume no longer exists in vServer. list_volume_usage reports the ids it could not measure in missing_volume_ids; use the restore points to size a deleted server's backups.

• Snapshot policies live in vServer, not here. list_snapshot_policies (and every server / volume snapshot operation) is exposed by vServer MCP Server. Snapshots are block-level and different from backups.