--- title: CLI and Web UI description: CLI flags and environment variables for inspecting local, Postgres, and Vercel Workflow runs. type: reference summary: Configure workflow inspect, workflow cancel, workflow web, workflow health, and observability tooling. related: - /docs/observability - /docs/configuration/worlds --- # CLI and Web UI The Workflow CLI uses flags first, then environment variables, then defaults or local inference. Vercel project and auth settings can often be inferred from `.vercel/project.json` and your Vercel CLI login. ## Target backend ### `--backend` / `-b` * Environment variable: `WORKFLOW_TARGET_WORLD` * Default: `local` * Backend to inspect: `local`, `vercel`, or a World package. ### `--authToken` / `-a` * Environment variable: `WORKFLOW_VERCEL_AUTH_TOKEN` * Default: Vercel CLI login * Vercel token for `--backend vercel`. ### `--project` * Environment variable: `WORKFLOW_VERCEL_PROJECT` * Default: inferred when possible * Vercel project ID for `--backend vercel`. ### `--team` * Environment variable: `WORKFLOW_VERCEL_TEAM` * Default: inferred when possible * Vercel team ID for `--backend vercel`. ### `WORKFLOW_VERCEL_PROJECT_NAME` * CLI flag: none * Default: inferred when possible * Vercel project slug used for dashboard links. ### `--env` / `-e` * Environment variable: `WORKFLOW_VERCEL_ENV` * Default: `production` * Vercel environment for `--backend vercel`. * Accepts `production` or `preview`. ## Web UI ### `--web` / `-w` * Environment variable: none * Default: disabled * Opens the relevant dashboard or web user interface (UI) instead of printing terminal output. ### `--webPort` * Environment variable: `WORKFLOW_WEB_PORT` * Default: `3456` * Port for the local web UI server. ### `--noBrowser` * Environment variable: `WORKFLOW_DISABLE_BROWSER_OPEN` * Default: browser opens * Prevents the CLI from opening a browser for web UI commands. ### `--localUi` * Environment variable: `WORKFLOW_LOCAL_UI` * Default: disabled * Uses the local web UI instead of the Vercel dashboard when inspecting Vercel. ### `--url` * Environment variable: none * Default: disabled * Prints the dashboard or run deep-link URL instead of opening a browser or starting a local server. ## Output and filtering ### `--json` / `-j` * Environment variable: none * Default: disabled * Prints machine-readable JSON where the command supports it. ### `--sort` * Environment variable: none * Default: `desc` for time-ordered listings; `workflow inspect attributes` orders keys alphabetically unless you pass this flag * Sort order for list commands. Accepts `asc` or `desc`. ### `--limit` * Environment variable: none * Default: `20` * Number of items returned per page for list commands. `workflow cancel` bounds it separately; see [`--limit` (cancel)](#--limit-cancel). ### `--cursor` * Environment variable: none * Default: unset * Pagination cursor for list commands. ### `--interactive` / `-i` * Environment variable: none * Default: disabled * Enables keyboard-controlled pagination for supported list commands. ## Inspect filtering Flags for `workflow inspect`. Each list view accepts a different subset, noted per flag. ### `--runId` / `-r` * Command: `workflow inspect` * Default: unset * Scopes the listing to one run. Required for `steps`, `events`, and `sleeps`; optional for `hooks`. Must be a run ID: `wrun_` followed by a 26-character ULID. ### `--limit` (inspect) * Command: `workflow inspect` * Default: `20` * Must be an integer between 1 and 100, the smallest page any inspect listing accepts. Larger pages are reachable by paging: pass `--cursor`, or `--interactive` to walk them. ### `--stepId` / `-s` * Command: `workflow inspect events` * Default: unset * Filters events to one step. ### `--hookId` * Command: `workflow inspect events` * Default: unset * Filters events to one hook. ### `--attribute` * Command: `workflow inspect runs` * Default: unset * Filters runs to those whose [attributes](/docs/observability/attributes) match every `key=value` pair given. Repeatable up to 8 times, and splits on the first `=` so a value may contain one. * Requires a backend with the analytics read path; ignored with a warning otherwise. * Cannot be combined with `--url` or `--web`, which hand off to the dashboard, or with `--withData`, which reads payloads from storage. Storage carries no attribute index. * Use `workflow inspect attributes` to discover which keys exist. ### `--since` / `--until` * Command: `workflow inspect runs`, `workflow inspect attributes` * Default: the backend's own window * Bounds the listing to a window. `--since` opens the window and accepts a relative duration (`30m`, `12h`, `7d`, `2w`) or a timestamp. `--until` is optional and defaults to now, so `--until` on its own is rejected. * Requires a backend with the analytics read path; ignored with a warning otherwise. ### `--withData` / `-d` * Command: `workflow inspect` * Default: disabled * Includes full input and output payloads in list views. Deprecated for list views — use `workflow inspect ` to read one item's payloads. Setting it also moves the read off the analytics path, which carries metadata only. ### `--decrypt` * Command: `workflow inspect` * Default: disabled * Decrypts encrypted values. Triggers an audit-logged key retrieval. ## Bulk cancel `workflow cancel ` cancels one run. Given a filter instead, it bulk-cancels a batch; bulk mode requires `--status` or `--workflowName`. ### `--status` * Command: `workflow cancel` * Default: unset * Restricts the batch to this status. Only `pending` and `running` are accepted; terminal runs cannot be canceled. * Also filters `workflow inspect runs`, which accepts any run status. It does not narrow `workflow inspect attributes`, which indexes keys per tenant rather than per run; passing it there warns and lists every key. ### `--workflowName` / `-n` * Command: `workflow cancel` * Default: unset * Restricts the batch to one workflow. Expects the generated workflow ID from `workflow inspect runs`, not the short function name. * Also filters `workflow inspect runs` and `workflow inspect attributes`. ### `--limit` (cancel) * Command: `workflow cancel` * Default: `50` * Maximum runs to cancel in one batch (1–100), the largest page the run listing serves. Only one batch is canceled per invocation; run the command again to cancel the next batch. ### `--confirm` / `-y` * Command: `workflow cancel` * Default: disabled * Skips the interactive confirmation prompt. ## Health checks ### `--port` / `-p` * Command: `workflow health` * Environment variable: `WORKFLOW_LOCAL_BASE_URL`, then `PORT` * Default: `3000` when neither env var is set * Local server port for health checks. * The flag writes `WORKFLOW_LOCAL_BASE_URL`. ## Verbose logging and updates ### `--verbose` / `-v` * Environment variable: `DEBUG` * Default: disabled * Enables verbose CLI output. ### `WORKFLOW_NO_UPDATE_CHECK` * CLI flag: none * Default: update check enabled * Set `1` to disable the CLI update check. ## Observability paths ### `WORKFLOW_OBSERVABILITY_CWD` * CLI flag: none * Default: `process.cwd()` * Working directory used by observability tooling to find `.vercel`, `.workflow-data`, and manifests. ### `WORKFLOW_MANIFEST_PATH` * CLI flag: none * Default: inferred * Explicit path to the workflow manifest for the web UI. --- For a semantic overview of all documentation, see [/sitemap.md](/sitemap.md) For an index of all available documentation, see [/llms.txt](/llms.txt) For agent-facing discovery, including API and MCP surfaces, see [/agents.md](/agents.md)