In the producing workspace, ensure outputs are defined in the Terraform configuration; they are automatically available to other workspaces in the same HCP Terraform organization.
In the consuming workspace's Terraform configuration, declare a tfe_outputs data source block with the organization and workspace name of the producing workspace.
Reference specific output values with data.tfe_outputs.<name>.values.<output_name> in the consuming module; the data source reads the current state outputs of the producing workspace.
Ensure the token used by the consuming workspace has read-state permission on the producing workspace; use a team API token with the appropriate permissions rather than a personal token for CI.
Avoid using terraform_remote_state for cross-workspace sharing in HCP Terraform; tfe_outputs is the recommended approach and avoids exposing the full state file to the consuming workspace.
Configure a run trigger from the producing workspace to the consuming workspace so the consumer re-plans automatically when the producer's outputs change.
Known gotchas
tfe_outputs only reads outputs that are not marked sensitive=true in the producing workspace unless the consuming workspace has explicit permission; sensitive outputs appear as null without an error.
The data source reads the last successful apply state; if the producing workspace has a pending or failed apply, the consuming workspace reads stale output values.
Cross-organization state sharing is not supported by tfe_outputs; both workspaces must be in the same HCP Terraform organization.
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