Recent releases from organizations and products in the Database category.
Hidehiko ASAHI shares a month of WalShadow private-preview testing, comparing replication latency with ClickPipes and checking data correctness, schema changes, and recovery.
ClickHouse 26.10 introduces function composition, more flexible joins, faster GROUP BY queries, disk spilling for IN subqueries, an experimental columns cache, and Freeform and SQLite formats.
supabase-flutter v3 adds type safety across queries, functions, realtime, and auth, and changes v2 behavior for anyone who upgrades. Nothing changes until you upgrade, and the release points to a migration guide for changes the compiler misses.
Neon3AI Gateway prepaid credits can now be purchased on the Free plan, starting at $5, and deleted projects can be recovered from the Neon Console for 7 days after deletion. Computes up to 16 CU, including autoscaling computes, now receive compute updates automatically instead of requiring manual restarts.
Neon Function Registry collects ready-made functions that agents install with neon functions new, starting from working code instead of reading docs and picking package versions. Templates cover Vercel Chat SDK, Mastra, TanStack AI, Context.dev, plus Postgres, AI Gateway, Object Storage, and Auth examples, and require Neon CLI 8.11.0 or later.
Production fleet stats show the compute node's local cache answers about 99.46% of page accesses and pageserver NVMe another 0.54%, leaving only about 39 requests per second to reach object storage. Object storage reads come almost entirely from branches or restores at old points in time and projects waking after scaling to zero, and only their first reads pay that cost.
The Tinybird Local image is about a third smaller to download and uses less memory when idle.
Workflows in GitHub Actions and Depot CI can now exchange an OIDC token for a short-lived PlanetScale service token instead of storing long-lived credentials in CI secrets. Organization admins define a trust policy in the dashboard specifying required claims, granted access, and token lifetime, and the planetscale/auth-action handles the exchange with the id-token: write permission.
The PlanetScale MCP server now runs on a new platform and sends database queries from four dedicated IP addresses. Postgres databases with IP restrictions must allow all four addresses for MCP queries to connect; Vitess databases need no changes.
Version 1.18.0 of the PlanetScale Terraform provider lets you set VTGate parameters in planetscale_vitess_branch and VTTablet and MySQL parameters in planetscale_vitess_keyspace.
Promoting a standby doesn't stop the old Postgres primary from taking writes. Here's how the control plane behind ClickHouse Managed Postgres fences it on bare metal, AWS and GCP, and why the design is fail-open.
The MongoDB CDC connector for ClickPipes is now generally available, with parallel snapshotting, structured ingestion into typed columns, lag and resource metrics, and Terraform and OpenAPI support.
Neon1Neon's MCP server now supports the stateless MCP spec revision dated 2026-07-28, commonly called MCP 2.0, and the old Server-Sent Events endpoints are retired. Tool calls on the server have grown about 14% week over week since August, and the MCP server now handles more tool calls and roughly twice the daily active users of the Neon CLI.
Supabase CLI v3 now targets the local stack by default and asks for confirmation before changing a hosted project. Deprecated commands are removed, and CI should be updated or pinned to v2 before upgrading.
The PlanetScale MCP server gains planetscale_submit_feedback for reporting wrong results, confusing errors, or missing tools, and planetscale_ask_question for recording questions the docs could not answer. Each submission includes the tool, organization, and database it concerns along with the user's ID and email so the team can follow up.
The September and October ClickStack releases bring ad hoc dashboard filters, PromQL for every chart type, ClickStack query tracing, a Trace logs tab, alerts from the chart explorer, and a long list of search page and collector improvements.
Emergency backups for Vitess databases copy data directly from the running database instead of the previous backup, so a fresh backup can be taken even when regular backups are failing. Production branches can copy from the primary or from a replica to reduce primary load, and backups are created from the dashboard, the API, or the CLI with pscale backup create --emergency.
clickhousectl v0.5.0 takes UDFs from local to Cloud, adds ClickHouse Managed Postgres observability, standardises the command surface, and lists the API-key permissions every Cloud command needs.
