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 |
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