Recently closed the Classic migration window and shipped the JSON data type to general availability.
Classic's migration window has closed. Automatic migration ran, but Free and Developer Workspaces that hit blocking issues need to migrate by October 16, 2026. After the grace period, remaining Free plan Classic Workspaces are deleted; Developer Workspaces are not. tb migrate-to-forward walks the process.
JSON columns are no longer a private beta. Every workspace can create and query JSON columns with no flag — dynamic and semi-structured event data is now fair game.
Query timeouts are tiered by plan: 10 seconds on Free, 20 on paid shared infrastructure, 30 on dedicated. Existing custom overrides are unchanged.
MCP tools gained output and planning control. execute_query and generated Pipe Endpoint tools accept response_format (CSVWithNames, JSON, or Markdown) and explain=true with optional settings to inspect a plan before running it.
Scheduled Copy Pipes can run on on-demand compute. Add ON_DEMAND_COMPUTE true to the Copy node on a paid Workspace to keep scheduled copies off your Workspace compute.
Deployments got cheaper and more forgiving.
- Changing the joined-table side of a Materialized View no longer rebuilds the view and everything downstream.
- Quarantine tables survive compatible deployments that change sorting keys, partition keys, or column types.
- Conditional TTL rules are supported in Data Source deployments.
Breaking: tinybird.rate_limits_log is now tinybird.cluster_control_log, with the same schema and six-month retention. Cluster protection events are visible on Overview and Observability, and failed Sink jobs now surface in sinks_ops_log.
Also: v1 local file ingestion is the CLI default, remote CSV, NDJSON, and Parquet files can be appended or replaced by URL, and Tinybird Local on arm64 requires a 4KiB page-size kernel.
