dbt Labs
The dbt Developer Agent shipped in Preview, bringing a project-aware coding agent to analytics engineering workflows for the first time.1
An analytics-aware coding agent arrived. The dbt Developer Agent reads your project graph, understands lineage, and drafts coordinated changes across SQL, YAML, tests, and docs in one pass. It's available in Preview for dbt platform customers with Copilot enabled. Agent Skills — the open-source repository of analytics engineering best practices for generalist coding agents — went GA alongside it.2
The dbt language gained several new capabilities in v1.12. The beta ships on_error: continue to allow downstream models to run when an upstream fails, a vars.yml file for project-level variable declarations, a selector selector method for reusing named selectors, JavaScript UDF support, overloaded UDFs, and dbt run-operation --sql for ad-hoc SQL execution. The new semantic layer YAML spec embeds semantic model definitions directly into model YAML entries — no more managing separate files.3
Upgrading to Fusion became self-serve. Platform customers can now migrate projects one at a time from the UI without filing a ticket. DuckDB support (Beta) lets teams run local dbt projects without a warehouse. Fusion release tracks — Nightly, Stable, Extended, and Fallback — give control over update cadence; new projects default to Fusion Stable.4
The Remote MCP Server added OAuth and Admin API access. AI tools like Claude and ChatGPT can now connect using existing dbt credentials. Two new tool sets — Admin API calls and search_product_docs / get_product_doc_pages — landed in the server. The 1.11.x series continued active maintenance through v1.11.11.5