Fivetran and dbt Labs are making dbt v2 and dbt State generally available. In addition, the merged company is introducing the Fivetran Context Layer, new dbt Wizard variants, and dbt Charts. All of these are focused on creating an open, vendor-neutral data infrastructure for AI agents.
dbt v2 is a complete rewrite of the dbt engine in Rust. According to the company, a project with 10,000 models runs up to ten times faster than v1. With the new release, the distinction between Core and Fusion is eliminated. dbt Core v1 remains the Python implementation, while Fusion becomes dbt v2. Both versions continue to be licensed under the Apache 2.0 license. The Fusion engine was already introduced last year, along with the first AI agents for data teams.
dbt State uses warehouse metadata and model SQL to determine what has changed and then builds, clones, or skips runs. Virgin Media O2 reports a 25 percent reduction in turnaround time and BigQuery compute costs. RxBenefits cites a 59 percent reduction in warehouse costs for scheduled jobs.
Open lakehouse and context for agents
Lake Compute, now in private beta, is a single-node SQL engine based on DuckDB that runs dbt models directly against Apache Iceberg tables. The Fivetran Context Layer (private beta) aggregates data and metadata for LLMs and agents via the open Agents Schema standard. Access is available through MCP, AI marketplaces such as Anthropic’s, and a ChatGPT plugin, among other channels.
“What we’re delivering now is the open infrastructure to put that context to work across systems,” says Anjan Kundavaram, Chief Product Officer.
dbt Wizard is coming to the terminal (CLI, public beta) and the desktop (private beta). dbt Charts, in public beta, defines BI visualizations as YAML alongside the models they rely on.