Skip to content

wfw run

Execute a workflow with full context aggregation.

Usage

wfw run <name> [options] [-- <input>...]

Arguments

Argument Description Required
<name> Workflow name Yes
-- <input>... Input files/directories (after --) No

Input Options

Primary documents to be analyzed or transformed.

Option Short Description
--input <path>... -in Add input files/directories (multiple allowed)

Context Options

Background materials and references.

Option Short Description
--context <path>... -cx Add context files/directories (multiple allowed)
--depends-on <name>... -dp Include outputs from other workflows

Model Options

Option Short Description
--provider <name> API provider: anthropic, openai
--profile <tier> Model tier: fast, balanced, deep
--model <model> -m Explicit model override (bypasses profile)

Thinking & Effort Options

Option Description
--enable-thinking Enable extended thinking mode
--disable-thinking Disable extended thinking (default)
--thinking-budget <num> Token budget for thinking (min 1024)
--effort <level> Effort level: low, medium, high (Opus 4.5 only)

API Options

Option Short Description
--temperature <temp> -t Override temperature (0.0-1.0)
--max-tokens <num> Override max tokens
--system <list> -p Comma-separated prompt names
--format <ext> -f Output format (md, txt, json, etc.)
--enable-citations Enable Anthropic citations support
--disable-citations Disable citations (default)

Output Options

Option Short Description
--export <path> -ex Copy output to external path

Execution Options

Option Short Description
--stream -s Stream output in real-time
--count-tokens Show token estimation only
--dry-run -n Save API request files and inspect in editor
--no-auto-deps Skip automatic execution of stale dependencies
--help -h Quick help

Output Location

  • Primary: .workflow/run/<name>/output.<format>
  • Hardlink: .workflow/output/<name>.<format>
  • Backups: .workflow/run/<name>/output-TIMESTAMP.<format>

Path Handling

  • Relative paths: Resolved from current working directory
  • Absolute paths: Used directly (access files anywhere on filesystem)
  • Directory paths: Expanded non-recursively to supported files
  • Glob patterns: Shell-expanded before passing to WireFlow (quote to prevent early expansion)

Duplicate paths are deduplicated. Inputs take precedence over context if the same path appears in both.

For batch processing, use wfw batch instead.

Examples

# Basic execution
wfw run 01-analysis --stream

# With model options
wfw run 01-analysis --profile deep --enable-thinking
wfw run 01-analysis --model claude-opus-4-5 --effort medium

# With input and context (multiple files)
wfw run report -in data.csv results.json -cx notes.md reference.pdf

# With dependencies
wfw run 02-synthesis -dp 01-analysis
wfw run 03-final -dp 01-analysis 02-synthesis

# Token estimation only
wfw run analysis --count-tokens

# Export to external path
wfw run analysis -ex ~/output/analysis.md

# Positional inputs after --
wfw run analysis -- reports/*.pdf

# Skip automatic dependency execution
wfw run 03-synthesis --no-auto-deps

See Also