What PlanetScale shipped

The MCP server moved to a new platform and now sends database queries from four dedicated IP addresses; Postgres databases with IP restrictions must allow all four, while Vitess needs no changes.

Vitess databases can now create emergency backups that copy directly from a running database, with production branches able to use the primary or a replica.

The MCP server now includes two tools for reporting problems and recording unanswered questions, sending a short note with the related tool, organization, and database to the PlanetScale team.

Terraform provider v1.16.0 adds Postgres storage configuration to the planetscale_postgres_branch resource, covering min and max storage, autoscaling, IOPS, and throughput.

Postgres databases now restart during a configurable maintenance window. Admins are emailed 7 days and 3 days ahead, and can move the window up to 3 times, via dashboard or API.

Email notifications now alert you when higher than usual public network egress is detected on Postgres and Neki databases.

Service tokens now display the last used IP, its location, and the network operator in the dashboard, API, and CLI. Audit logs also include the network organization.

Beyond MoveTables, Vitess branches now support Materialize, LookupVindex, VDiff, and PlannedReparentShard through pscale branch vtctl (alias vtctld), plus helpers for listing tablets and updating shard tablet controls.

The Vitess branch Workflows page now lists Materialize workflows alongside MoveTables, showing stream states, replication lag, and materialized tables. Creating and managing them stays in the CLI.

Live connections to Neki routers are now visible in the dashboard and via CLI commands, and running queries can be cancelled or sessions terminated. Only router sessions are shown, not router-to-shard connections.

Vitess keyspace disks can now be set to grow automatically up to a limit, stay off, or shrink to a chosen size, from the dashboard, API, or CLI. Billing follows disk capacity, not the limit.

The Heroku Migrator now drops secondary indexes during the initial copy and rebuilds them after data lands, and migrates all schemas rather than just public. A Verify Migration option compares the two databases.

The patroni.slots cluster parameter is gone; Postgres logical replication slots are now detected automatically. Failover support should still be enabled on slots to keep them through switchovers.

Metal sizes M-10 through M-160 are now available for Neki on AWS, matching Postgres NVMe sizes. Storage capacity is chosen separately from vCPU and memory and stays fixed.

Neki now supports read-only replicas, up to 8 per shard, that serve reads and never become primary. Reads are routed to them via the Connect page or the __neki.target setting.

Huge pages can now be enabled for Postgres and Neki on supported cluster sizes, set via huge_pages in the Parameters tab. Applying the change requires a restart.

Terraform provider v1.13.0 lets you set and read Neki router parameters on planetscale_neki_router, and the planetscale_databases data source can now filter by kind. Postgres bouncer users on Terraform need v1.14.0 or later.

CLI v0.346.0 shows a Requires update column in parameter listings for Postgres branches and Neki profiles, admins, and sidecars, with a matching requires_update field in JSON output.

Azure is now a cloud provider option for creating Postgres and Neki databases, with the same features as other clouds. Vitess databases are not yet available on Azure.

A single deploy request can now opt into aggressive cutover through the CLI and API, and can be changed before the schema change is deployed.

Vitess deploy requests can now retry only the tables in a failed migration instead of restarting the whole deploy. Healthy tables keep their progress. Available via pscale deploy-request retry.

The MCP server can now pull branch metrics for Vitess, Postgres, and Neki databases, summarizing CPU, memory, disk, replication lag, and connections from the last hour by default.

Organization administrators can limit MCP query access to none, read-only, or read and write. The setting applies the next time a client is authorized; existing clients keep current access.

Postgres now supports dedicated read replicas that can be configured independently from the primary cluster, in the same or another region, via dashboard, API, CLI, or Terraform.

Two Prometheus metrics now track planned and emergency reparents, with labels for result and for the component that ran the reparent.

The CLI now includes pscale keyspace parameters for viewing and changing VTTablet and MySQL parameters on a Vitess keyspace, including listing current and default values, tracking rollout changes, and resetting parameters.

Schema recommendation webhooks now fire for Postgres databases when a new recommendation is created, and a custom Authorization header can be set on deliveries.

Organization invitations now let you choose a member's role before sending, so new members join with that role instead of defaulting to Organization Member.