Workspace settings control how the workspace behaves and what defaults apply to every member. Open settings by clicking the gear icon in the workspace header.
General settings
| Setting | Description | Who can edit |
|---|
| Workspace name | Display name in dashboard and sidebar | Admin, Owner |
| Workspace slug | URL identifier (read-only after creation) | — |
| Delete workspace | Permanently removes workspace and all tasks | Owner only |
Deleting a workspace is permanent and cannot be undone.
Configuration hierarchy
Settings follow a three-level hierarchy, where each level can override the one below:
| Priority | Scope | Set by |
|---|
| Highest | Workspace user overrides | Individual member (per-workspace) |
| Middle | User defaults | Individual member (all workspaces) |
| Lowest | Workspace defaults | Workspace admin |
Members can save workspace overrides as new user defaults using the Save to user defaults button.
Configuration sections
The workspace defaults tab includes:
| Section | Description |
|---|
| Instructions | CLAUDE.md content appended to agent sessions |
| Default Model | AI model for new tasks |
| Permission Mode | How the agent requests approval |
| Auto-Sleep | Idle timeout before task suspension |
| Environment Variables | Workspace-level secrets and config values |
| MCP Servers | Model Context Protocol server connections |
| Plugins | Enabled plugins from the catalog |
| Initialization Script | Bash script that runs on container start |
| Shell Configuration | Custom .bashrc content for container shells |
Default AI provider and model
Admins set a default AI provider and model for all tasks. Members can override in their personal settings.
Permission mode options:
| Mode | Behavior |
|---|
| Use user preference | Each member’s personal setting applies |
| Default | Standard permission prompts |
| Accept Edits | Auto-approve file edits, prompt for other actions |
| Plan Mode | Agent creates plans without executing |
| Bypass Permissions | Auto-approve all actions |
Workspace instructions
Add custom instructions appended to every agent session. These augment (don’t replace) each member’s personal instructions.
Auto-sleep
Configure idle timeout before Whim suspends a task (disabled, or 1 minute to 1 hour). Suspended tasks resume instantly.
Members
The Members tab controls workspace access.
| Role | Capabilities |
|---|
| Viewer | View tasks and settings (read-only) |
| Editor | Create and run tasks, edit personal workspace settings |
| Admin | All editor capabilities plus: manage members, edit workspace defaults, manage images |
Environment variables
Environment variables inject API keys, database URLs, and other configuration into every task container.
Scopes and precedence
| Priority | Scope | Applies to |
|---|
| Highest | Workspace user overrides | Your tasks in this workspace |
| Middle | User defaults | Your tasks in all workspaces |
| Lowest | Workspace defaults | All members in this workspace |
Adding variables
- Workspace-level — Workspace Settings > Workspace Defaults > Environment Variables
- User-level — Settings > User Defaults > Environment Variables
- Per-workspace overrides — Settings > My Workspace Overrides > Environment Variables
Multi-file support
Variables are organized into .env files. You can create multiple files following .env / .env.<suffix> naming (e.g., .env.local, .env.production). Files are processed in order: .env first, then additional files alphabetically. Later files override matching keys.
Secret masking
Whim auto-detects likely secrets based on key names (containing secret, password, token, or matching patterns like API_KEY, PRIVATE_KEY) and masks their values in the UI. Toggle the secret status of any variable with the lock icon.
Secrets are masked for display. They’re stored encrypted at rest and injected securely into containers.
Disabling workspace variables
Members can disable specific workspace-level variables without removing them. Disabled variables are filtered out before merging and won’t appear in your task containers.
How variables are injected
When a task starts, Whim resolves variables from all three levels and injects them:
- As environment variables — merged into the container’s environment (
process.env)
- As
.env files — written to /app inside the container for framework auto-discovery
System variables
Whim also injects system variables that cannot be overridden:
| Variable | Description |
|---|
WHIM_AI_PROVIDER | Configured AI provider |
WHIM_AI_RUNTIME | AI runtime in use |
WHIM_TASK_ID | Current task identifier |
WHIM_WORKSPACE_ID | Workspace identifier |
WHIM_TEAM_ID | Team identifier |
WHIM_GITHUB_TOKEN | GitHub access token for Git operations |
WHIM_CLONE_URL | Repository clone URL |
