2026 W36

Bruin Cloud

Dashboards

• New Version history and restore → Browse earlier draft and published versions, then restore a checkpoint into the current draft. Review it before publishing; restoring doesn't replace the live dashboard.
• New Vega-Lite widgets → Build custom visualizations beyond the built-in chart types, including layered and faceted charts. Available in both Cloud and DAC.
• New Dual y-axes → Put measures with different scales, such as revenue and conversion rate, on separate axes in supported charts. DAC supports the same setup with y2.
• New Send filters to the agent → Attach the dashboard's active filters to a chat message so the agent has the context for the slice you're viewing.
• Improved Charts inside dashboard chat → Agent-produced charts render directly in the conversation.

Agents and streams

• New File inputs in Teams and Telegram → Send files to Bruin in either chat integration and include them in the agent's analysis.
• New Repository-defined skills → Keep reusable agent instructions alongside your repository content and make them available to agent runs automatically.
• New Personal API tokens → Create tokens under your own identity, scoped to a team and selected permissions, with an expiry you choose.
• New Multi-table streams → Select several source tables in the Streams wizard and replicate them in one stream. Custom destination table names remain available for single-table streams.

Bruin CLI

• New Relationship checks → Add a relationships check to a column with foreign_key metadata to find keys missing from its parent asset. Add a separate not_null check if nulls should fail, and declare depends when the parent must refresh first.
• New GCS sensors → Wait for an object, a matching path, or a prefix in Google Cloud Storage before running downstream work. Choose one-shot, polling, or skip behavior.
• New PostHog analytics starter → Run bruin init posthog-bigquery for identity resolution, account rollups, engagement scoring, retention reports, and a DAC dashboard.
• Improved Cloud automation controls → Manage connection sets and agent memory, export agent threads, and trigger scheduled agents from the CLI. Use bruin cloud config set-team to save your default team rather than passing it on every command.
• Improved Choose the dashboard definition to retrieve → Use bruin cloud dashboards get --state draft|published to fetch the version your workflow needs. Draft access requires edit permission.

Bruin VS Code

• New Custom MCP clients → Connect local Bruin or Bruin Cloud to a client that uses a JSON mcpServers configuration. Add its config path in Custom clients and enable the connection.
• New Start a query from a database → Use the + action in the Databases panel to open an empty SQL query with its connection and environment already selected.
• New Persistent full refresh for ingestion assets → Use the asset panel's Full Refresh switch to save parameters.full_refresh: true, instead of remembering a run-time flag each time.
• Fix Asset panels refresh when switching files, including on Windows, rather than staying on the previous asset until reopened.

DAC

• New Heatmap labels and color scales → Show values inside cells and define multi-color scales with fixed ranges, so colors retain their meaning when filters change.
• New Slice labels and colors → Use a slices map on pie, treemap, and funnel widgets to rename categories or give them consistent colors without changing the underlying data.
• Improved Per-series styling → Customize individual line, area, and combo-chart series with colors, curves, dash patterns, and point markers.

ingestr

• New New sources → Okta, Deel, Cloudflare Radar, RIPEstat, Sklik, 2Checkout, SatisMeter, ABRA Flexi, Twenty CRM, Sumble, and Vertica.
• New New destinations → Load into Vertica or an Iceberg catalog on Cloudflare R2, and send profiles or events to CleverTap. CleverTap profiles upsert; events append, so rerunning the same event rows creates duplicates.
• New Select a subset of CDC tables → Pass a comma-separated --source-table list to replicate several tables without selecting every eligible table. Changing the selection starts a new snapshot; clean up obsolete PostgreSQL replication slots after switching.
• New Trace rows to a load → Bounded ingestion runs add _ingestr_run_id, so rows from the same run can be identified across tables. Use --no-run-id if you don't want the extra column.
• Improved Large HTTP file ingestion → CSV, JSON, and JSONL loads can start delivering batches while the file downloads, including gzip responses. Interrupted downloads can resume within the same run when the server supports ranges and a stable file validator.
• New DuckLake table layout → Set --partition-by and --cluster-by for single-table DuckLake loads to control partitioning and physical sort order. Requires a compatible DuckLake extension.
• Fix PostHog event extraction now pages through large date ranges without silently stopping at a partial result.

Bruin Academy

• New Ask the Data → Learn SQL alongside AI-assisted analysis in the beginner course, including how to check and correct agent-generated queries against a local dataset.