Skip to main content
September 22, 2026 Every DinoAI agent ran against the workspace’s default Agent environment. An agent can now name an environment of its own, and run on that environment’s configuration instead.
  • In the builder. The agent builder has an Environment section. Leave it on Workspace default and nothing changes. Pick a named environment and the agent is pinned to it.
  • In YAML. Agents defined in your repo take an optional environment: <slug> field, and the builder round-trips it, so the form and the YAML view stay in agreement.
What the environment decides The agent uses that environment’s warehouse connection, its dbt™ version, and its partial parsing setting, and its SQL tools run through the same connection. So an agent can be held on dbt™ 1.12 while the rest of the workspace stays on 1.11, or point at a warehouse role scoped for agent work. The environment is resolved and pinned before the agent’s runtime starts, which is what makes the dbt™ version take effect rather than being applied too late to matter. It works for YAML you wrote by hand and have not synced to the workspace. A session that has already started keeps its environment for follow-up turns, so editing the agent mid-run cannot move it. When it cannot be resolved An unknown slug fails the session at startup with an error naming the problem, rather than quietly falling back. This is deliberately stricter than model families, which do fall back and warn: an agent silently running against the wrong warehouse is worse than one that does not start. In the builder, an environment that has since been deleted blocks the deploy until you choose another. Remove the field and the agent goes back to the workspace’s default Agent environment. Agent environments also gained the dbt™ partial parsing toggle that Bolt environments already had, under Settings → Environments → Agent. See YAML configuration and Creating agents in the app. Shipped in Paradime v6.13.0.