Set up a Dagster+ organization and configure a Hybrid or Serverless agent; enable branch deployments in your CI/CD configuration so each pull request deploys an isolated code location
Add a GitHub Actions (or GitLab CI) workflow that calls the dagster-cloud CI GitHub Action or runs 'dagster-cloud branch-deployment create --branch-name $BRANCH_NAME' using a CI API token stored as a repository secret
The branch deployment creates an isolated code location in Dagster+ where the asset graph, sensors, and schedules from the PR branch are loaded without affecting the production deployment
Run asset definition validation in CI by executing 'dagster definitions validate' against the branch deployment; this command checks that all assets, jobs, sensors, and resources resolve without errors before the PR is merged
After the PR merges, the branch deployment is automatically deleted; configure cleanup behavior with the --delete-on-close flag in your CI workflow to ensure no stale code locations accumulate in Dagster+
Known gotchas
Branch deployments require Dagster+ (the managed cloud offering); OSS Dagster does not support branch deployments natively — teams on OSS must implement their own isolation strategy using separate Docker images or namespaces
The CI API token used for branch deployment creation has organization-level permissions; store it as an encrypted secret and avoid logging the token value in CI workflow output
Assets defined in a branch deployment share the same asset key namespace as production; if a branch introduces a key conflict (e.g., renaming an asset), materializing it from the branch deployment will update the shared asset event log and may confuse production monitoring
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp