Keyboard shortcuts

Press or to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Command-Line Options

🎛️ Complete reference for all CLI flags and options.

ReasonKit’s CLI is designed for power users and automation. Every option has both a short and long form.

Global Options

These options work with all commands:

ShortLongDefaultDescription
-p--profilebalancedReasoning profile to use
-o--outputprettyOutput format (pretty, json, markdown)
-v--verbosefalseEnable verbose output
-q--quietfalseSuppress all non-essential output
-h--help-Show help message
-V--version-Show version information
--config~/.config/reasonkit/config.tomlConfig file path
--no-colorfalseDisable colored output

Profile Selection

Quick Access

# Using --profile flag
rk-core think "question" --profile balanced

# Using shorthand flags
rk-core think "question" --quick      # ~10 seconds
rk-core think "question" --balanced   # ~20 seconds (default)
rk-core think "question" --deep       # ~1 minute
rk-core think "question" --paranoid   # ~2-3 minutes

Profile Options

FlagEquivalentTimeUse Case
--quick--profile quick~10sLow-stakes, time-sensitive
--balanced--profile balanced~20sMost decisions
--deep--profile deep~1mHigh-stakes decisions
--paranoid--profile paranoid~2-3mCritical, irreversible

Provider Options

Provider Selection

# Auto-detect (uses first available key)
rk-core think "question"

# Explicit provider
rk-core think "question" --provider anthropic
rk-core think "question" --provider openai
rk-core think "question" --provider openrouter
rk-core think "question" --provider ollama

Model Selection

# Use default model for provider
rk-core think "question" --provider anthropic

# Specify exact model
rk-core think "question" --model claude-sonnet-4-20250514
rk-core think "question" --model gpt-4-turbo
rk-core think "question" --model llama3.2

Provider-Specific Options

OptionDescriptionExample
--providerLLM provider to useanthropic, openai, openrouter, ollama
--modelSpecific model IDclaude-sonnet-4-20250514
--api-keyAPI key (overrides env)sk-...
--base-urlCustom API endpointhttp://localhost:11434

Output Options

Format Selection

# Pretty output (default, human-readable)
rk-core think "question" --output pretty

# JSON output (machine-readable)
rk-core think "question" --output json

# Markdown output (documentation-ready)
rk-core think "question" --output markdown

# Short alias
rk-core think "question" -o json

Output Control

OptionDescriptionExample
--output, -oOutput formatpretty, json, markdown
--no-colorDisable ANSI colorsFor piping/logging
--quiet, -qMinimal outputOnly final result
--verbose, -vDetailed outputInclude debug info

Output Redirection

# Save to file
rk-core think "question" --output json > analysis.json

# Pipe to jq
rk-core think "question" -o json | jq '.synthesis'

# Suppress progress, keep result
rk-core think "question" -q > result.txt

Timing Options

Timeout Control

# Default timeout (180 seconds)
rk-core think "question"

# Custom timeout
rk-core think "question" --timeout 300

# No timeout (wait indefinitely)
rk-core think "question" --timeout 0

Streaming

# Enable streaming (see output as it generates)
rk-core think "question" --stream

# Disable streaming (wait for complete response)
rk-core think "question" --no-stream

ThinkTool Options

Tool Selection

# Use specific tools only
rk-core think "question" --tools gigathink,laserlogic

# Exclude specific tools
rk-core think "question" --exclude proofguard

# Tool aliases
rk-core think "question" --tools gt,ll,br,pg,bh

Tool-Specific Settings

# GigaThink perspectives
rk-core think "question" --gigathink-perspectives 15

# ProofGuard source count
rk-core think "question" --proofguard-sources 5

# BrutalHonesty severity
rk-core think "question" --brutalhonesty-severity high
OptionValuesDefault
--gigathink-perspectives3-2010
--laserlogic-depthquick, standard, deepstandard
--proofguard-sources1-103
--brutalhonesty-severitylow, medium, highmedium

Input Options

Reading Input

# Direct question
rk-core think "Is this a good idea?"

# From stdin
echo "Should I do this?" | rk-core think -

# From file
rk-core think --file question.txt

# From clipboard (requires xclip/pbpaste)
rk-core think --clipboard

Context Addition

# Add context file
rk-core think "Should I merge this PR?" --context diff.txt

# Multiple context files
rk-core think "question" --context file1.txt --context file2.txt

# URL context (fetches and includes)
rk-core think "question" --context-url https://example.com/docs

Cache Options

# Enable caching (default)
rk-core think "question" --cache

# Disable caching
rk-core think "question" --no-cache

# Clear cache before running
rk-core think "question" --clear-cache

# Set cache TTL (seconds)
rk-core think "question" --cache-ttl 3600

Debug Options

# Show debug information
rk-core think "question" --debug

# Show API requests/responses
rk-core think "question" --debug-api

# Dry run (show what would be sent)
rk-core think "question" --dry-run

# Show timing breakdown
rk-core think "question" --timing

Combining Options

Options can be combined freely:

# Full example
rk-core think "Should I accept this job offer?" \
  --profile deep \
  --provider anthropic \
  --model claude-sonnet-4-20250514 \
  --output json \
  --timeout 300 \
  --proofguard-sources 5 \
  --verbose

# Shorthand version
rk-core think "Should I accept this job offer?" \
  --deep -o json -v --proofguard-sources 5

Option Precedence

Options are applied in this order (later overrides earlier):

  1. Built-in defaults
  2. Config file settings
  3. Environment variables
  4. Command-line flags
# Config says balanced, but CLI overrides to deep
rk-core think "question" --deep