Slash Commands

Quick shortcuts for every common action.

Overview

Slash commands are shortcuts you type directly into the Franklin prompt. They start with / and trigger built-in actions instantly. Type /help to see the full list at any time.

Session Commands

Manage your current conversation and workflow:

bash
/plan          # Create a structured plan before executing
/execute       # Execute the current plan step-by-step
/compact       # Compress conversation to save context window
/retry         # Retry the last failed action
/history       # List past sessions
/resume [id]   # Resume a previous session

Code Commands

Git, testing, and code management shortcuts:

bash
/commit        # Stage and commit changes with a generated message
/push          # Push current branch to remote
/pr            # Create a pull request
/review        # Review the current diff
/test          # Run the project's test suite
/fix           # Auto-fix linting and type errors
/search "q"    # Search across the codebase

/commit is smart

The /commitcommand analyzes your staged changes, generates a descriptive commit message, and asks for confirmation before committing. It follows your project's commit conventions automatically.

Info Commands

Check status, costs, and memory:

bash
/model [name]  # Show or switch the current model/profile
/wallet        # Show wallet address and balance
/cost          # Per-model spend breakdown for this session
/tokens        # Show token usage for the current conversation
/learnings     # View learned preferences
/brain         # Explore the knowledge graph
/help          # Show all available commands

Power Commands

Advanced commands for complex reasoning:

bash
/ultrathink    # Extended thinking mode — deeper reasoning
/ultraplan     # Comprehensive planning with multi-step breakdown

Extended thinking

/ultrathink enables extended thinking mode where Franklin spends more time reasoning before responding. Useful for complex debugging, architecture decisions, and math problems. It automatically selects a reasoning-optimized model.