PlanetScale
Force cutover shipped for deploy requests that stall on long-running transactions, giving teams an escape hatch without manual intervention.1
Deploy request reliability tightened. A "force cutover now" option appears when the cutover phase is delayed by long-running transactions, aggressively stopping them to unblock schema changes.2 An "aggressive cutover" setting makes this the default for all deploy requests on a database. A new API endpoint also lets automations check whether a target branch has enough storage before applying schema changes.3
Postgres observability expanded. Weekly database reports now email a recap for each Postgres database covering anomalies, most expensive queries (with p99 latency and cache hit ratio), Schema Recommendations, Traffic Control activity, storage delta, and daily CPU/memory averages.4 Opt-in per database in user settings.
Terraform can now create and delete databases. Applying a planetscale_postgres_branch or planetscale_vitess_branch resource creates the parent database if absent; destroying the last branch deletes it.5 Previously, the database lifecycle required out-of-band management.
Negated search in Query Insights lets engineers filter noise with !6 prefixes — !table:users or p99:>100 !table:sessions — on both Postgres and Vitess, making it faster to isolate problem queries.
Agentic workflows got dedicated MCP tooling. The insights-only MCP server at mcp.pscale.dev/mcp/planetscale-insights-only exposes Insights data and Schema Recommendations without query execution — designed for agents that should analyze but not mutate.7 The Go SDK v0.164.0 added a schema recommendation API.8
CLI and SDK housekeeping. Move table cancel updated to use async operations (v0.285.0).9 database-js v1.20.1 improved Cloudflare error detection in UnknownError messaging.10 The CLI Docker image now publishes as multi-arch (linux/amd64 + linux/arm64).