Skip to content

CLI Reference

Complete reference for all Claudio commands.

Global Options

-c, --config string   Config file path (default: ~/.config/claudio/config.yaml)
-h, --help            Help for command

Commands

claudio init

Initialize Claudio in the current git repository.

claudio init

Creates a .claudio/ directory for session state and worktrees.

Requirements: - Must be run in a git repository - Directory must be writable


claudio start

Start a new session and launch the TUI dashboard.

claudio start [session-name] [flags]

Arguments: | Argument | Description | |----------|-------------| | session-name | Optional name for the session |

Flags: | Flag | Description | |------|-------------| | --new | Force start a new session, replacing any existing one |

Examples:

# Start with default name
claudio start

# Start with custom name
claudio start my-feature

# Force new session (discard existing)
claudio start --new

Behavior: - If a previous session exists, prompts to recover or start fresh - Launches interactive TUI dashboard - Detects and offers to reconnect to running tmux sessions


claudio add

Add a new backend instance with a task.

claudio add [task description] [flags]

Arguments: | Argument | Description | |----------|-------------| | task description | Description of the task for the backend |

Flags: | Flag | Short | Description | |------|-------|-------------| | --start | -s | Automatically start the instance after adding | | --depends-on | -d | Comma-separated list of instance IDs or task names this task depends on |

Examples:

# Add and auto-start
claudio add "Implement user authentication" --start

# Add without starting
claudio add "Write unit tests for auth module"

# Quote complex tasks
claudio add "Refactor the User model to include email validation and password hashing"

# Task chaining - add with dependencies
claudio add "Write unit tests" --depends-on "abc123"
claudio add "Deploy to staging" -d "tests,build"

# Multiple dependencies
claudio add "Integration tests" --depends-on "unit-tests,api-setup"

Task Chaining:

When you specify --depends-on, the instance will wait in pending state until all its dependencies have completed. This enables:

  • Sequential execution where order matters
  • Building complex workflows with parallel and serial phases
  • Ensuring prerequisites are met before dependent tasks start

Dependencies can be specified by: - Instance ID (e.g., abc123) - Task name substring (e.g., tests matches "Write unit tests") - Comma-separated list for multiple dependencies


claudio status

Display current session status and all instances.

claudio status

Output includes: - Session name and state - List of all instances - Instance status (working, paused, completed, etc.) - Branch names - Basic metrics


claudio stop

Stop all running instances and optionally cleanup.

claudio stop [flags]

Flags: | Flag | Short | Description | |------|-------|-------------| | --force | -f | Force stop without prompts |

Behavior: - Prompts for what to do with each instance's work - Options: create PR, keep branch, discard - Use --force to skip prompts and stop immediately


claudio pr

Create a GitHub pull request for an instance.

claudio pr [instance-id] [flags]

Arguments: | Argument | Description | |----------|-------------| | instance-id | ID of the instance (optional if only one exists) |

Flags: | Flag | Short | Description | |------|-------|-------------| | --title | -t | Override PR title | | --body | -b | Override PR body | | --draft | -d | Create as draft PR | | --reviewer | -r | Add reviewers (repeatable) | | --label | -l | Add labels (repeatable) | | --closes | | Link issues to close | | --no-ai | | Skip AI generation | | --no-rebase | | Skip rebasing on main | | --no-push | | Don't push before creating PR |

Examples:

# Create PR with AI-generated content
claudio pr abc123

# Create draft PR
claudio pr abc123 --draft

# Override title
claudio pr abc123 --title "feat: add user auth"

# Add reviewers and labels
claudio pr abc123 --reviewer teammate1 --reviewer teammate2 --label enhancement

# Link to issue
claudio pr abc123 --closes 42

# Skip AI generation
claudio pr abc123 --no-ai


claudio config

View or modify configuration.

claudio config [command]

Subcommands:

claudio config (no args)

Opens interactive configuration UI.

claudio config show

Display current configuration non-interactively.

claudio config show

claudio config init

Create a default config file with comments.

claudio config init

claudio config set

Set a configuration value.

claudio config set <key> <value>

Examples:

claudio config set completion.default_action auto_pr
claudio config set branch.prefix "Iron-Ham"
claudio config set pr.draft true

claudio config edit

Open config file in $EDITOR.

claudio config edit

claudio config path

Show config file location.

claudio config path

claudio config reset

Reset configuration to defaults.

claudio config reset

claudio config theme

Manage color themes.

claudio config theme [command]

Subcommands: | Command | Description | |---------|-------------| | list | List all available themes (built-in and custom) | | create <name> | Generate a custom theme template | | export <theme> | Export a theme's YAML definition |

Examples:

# List all themes
claudio config theme list

# Create a custom theme
claudio config theme create my-theme

# Export for customization
claudio config theme export dracula > ~/.config/claudio/themes/my-dracula.yaml

Custom themes are stored in ~/.config/claudio/themes/ and automatically discovered.


claudio cleanup

Clean up stale worktrees, branches, and tmux sessions.

claudio cleanup [flags]

Flags: | Flag | Short | Description | |------|-------|-------------| | --dry-run | | Show what would be cleaned without making changes | | --force | -f | Skip confirmation prompt | | --worktrees | | Clean up only worktrees | | --branches | | Clean up only branches | | --tmux | | Clean up only tmux sessions | | --all-sessions | | Clean resources from all sessions, not just current | | --deep-clean | | Perform thorough cleanup including orphaned resources | | --foreground | | Run synchronously instead of in background | | --job-status | | Check status of a background cleanup job |

Background Execution:

By default, cleanup runs in the background. Resources are snapshotted at invocation time, so new worktrees created during cleanup are not affected.

Examples:

# Preview cleanup
claudio cleanup --dry-run

# Clean everything (runs in background)
claudio cleanup --force

# Run synchronously
claudio cleanup --force --foreground

# Check background job status
claudio cleanup --job-status abc123

# Deep clean all sessions
claudio cleanup --all-sessions --force --deep-clean

# Clean only worktrees
claudio cleanup --worktrees

What gets cleaned: - Worktrees in .claudio/worktrees/ with no active session - Branches matching <prefix>/* not associated with active work - Orphaned claudio-* tmux sessions - Background job files older than 24 hours (auto-cleaned)


claudio harvest

Review and commit work from completed instances.

claudio harvest [flags]

Flags: | Flag | Description | |------|-------------| | --all | Process all worktrees (commit + pr + cleanup) | | --commit | Auto-commit all uncommitted changes | | --pr | Create PRs for committed branches | | --cleanup | Remove worktrees with no changes |

Examples:

# Interactive harvest
claudio harvest

# Full automation
claudio harvest --all

# Just commit completed work
claudio harvest --commit

# Commit and create PRs
claudio harvest --commit --pr


claudio remove

Remove a specific instance and its worktree.

claudio remove <id> [flags]

Arguments: | Argument | Description | |----------|-------------| | id | Instance ID to remove |

Flags: | Flag | Short | Description | |------|-------|-------------| | --force | -f | Remove even with uncommitted changes |

Examples:

# Remove instance (prompts if uncommitted changes)
claudio remove abc123

# Force remove
claudio remove abc123 --force


claudio stats

Display resource usage and cost statistics.

claudio stats [flags]

Flags: | Flag | Description | |------|-------------| | --json | Output as JSON |

Output includes: - Total token usage (input/output) - Estimated API costs - Per-instance breakdown - Budget limit status

Examples:

# Human-readable output
claudio stats

# JSON for scripting
claudio stats --json


claudio sessions

Manage Claudio sessions.

claudio sessions [command]

Subcommands:

claudio sessions list

List recoverable sessions and orphaned tmux sessions.

claudio sessions list

claudio sessions attach

Attach to an existing session by ID.

claudio sessions attach <session-id>

Reconnects to an existing session and launches the TUI.

claudio sessions recover

Recover a previous session.

claudio sessions recover

Reconnects to running tmux sessions and restores the TUI.

claudio sessions clean

Clean up stale session data.

claudio sessions clean


claudio plan

Generate a structured task plan from an objective.

claudio plan [objective] [flags]

Arguments: | Argument | Description | |----------|-------------| | objective | The goal or feature to plan (optional, can be interactive) |

Flags: | Flag | Description | Default | |------|-------------|---------| | --dry-run | Show the plan without executing | false | | --output-format | Output format: json, issues, or both | issues | | --multi-pass | Use 3 independent strategies to generate plans | false | | --no-confirm | Skip confirmation prompt | false | | --output | Write plan JSON to a specific file path | .claudio-plan.json | | --labels | Labels to add to GitHub Issues (comma-separated) | - |

Examples:

# Basic planning
claudio plan "Add user authentication"

# Dry run to review plan
claudio plan --dry-run "Refactor the API layer"

# Output as GitHub issues
claudio plan --output-format issues "Implement caching"

# Multi-pass planning for complex tasks
claudio plan --multi-pass "Redesign the data model"

Output Formats: - json - Outputs a structured JSON plan file - issues - Creates GitHub issues for each task - both - Creates both JSON file and GitHub issues

See Plan Mode Guide for detailed documentation.


claudio ultraplan

Intelligent hierarchical planning with 4-phase parallel execution.

claudio ultraplan [objective] [flags]

Arguments: | Argument | Description | |----------|-------------| | objective | The goal or feature to implement |

Flags: | Flag | Description | Default | |------|-------------|---------| | --plan | Use existing plan file instead of planning phase | - | | --max-parallel | Maximum concurrent child sessions (0 = unlimited) | 3 | | --dry-run | Run planning only, output plan without executing | false | | --no-synthesis | Skip synthesis phase after execution | false | | --auto-approve | Auto-approve spawned tasks without confirmation | false | | --multi-pass | Use 3 competing strategies, then select best | false | | --review | Always open plan editor before execution | false |

Examples:

# Basic ultraplan
claudio ultraplan "Implement OAuth2 authentication"

# Increase parallelism
claudio ultraplan --max-parallel 5 "Add comprehensive tests"

# Dry run to review plan
claudio ultraplan --dry-run "Refactor to microservices"

# Use existing plan file
claudio ultraplan --plan my-plan.json

# Multi-pass for complex architecture
claudio ultraplan --multi-pass "Redesign the authentication system"

# Skip synthesis for manual review
claudio ultraplan --no-synthesis "Update deprecated APIs"

Phases: 1. Planning - The backend explores the codebase and generates a structured plan 2. Context Refresh - Review and approve the generated plan 3. Execution - Child instances execute tasks in parallel (respecting dependencies) 4. Synthesis - Coordinator reviews all outputs and identifies integration issues

See Ultra-Plan Guide for detailed documentation.


claudio adversarial

Iterative implementation with reviewer feedback loop.

claudio adversarial [task] [flags]

Arguments: | Argument | Description | |----------|-------------| | task | The task to implement with review cycles |

Flags: | Flag | Description | Default | |------|-------------|---------| | --max-iterations | Maximum implement-review cycles (0 = unlimited) | 10 | | --min-passing-score | Minimum score (1-10) required for approval | 8 |

Examples:

# Basic adversarial review
claudio adversarial "Implement user authentication with JWT"

# Limit review cycles
claudio adversarial --max-iterations 5 "Refactor the API layer"

# Strict quality requirements
claudio adversarial --min-passing-score 9 "Implement encryption module"

# Combined for critical code
claudio adversarial --max-iterations 5 --min-passing-score 9 "Implement auth tokens"

How it works: 1. An IMPLEMENTER instance works on the task 2. When ready, submits work for review 3. A REVIEWER instance examines the code and provides feedback 4. Loop continues until approved with passing score, or max iterations reached

See Adversarial Review Guide for detailed documentation.


claudio validate

Validate ultraplan JSON files before execution.

claudio validate <plan-file> [flags]

Arguments: | Argument | Description | |----------|-------------| | plan-file | Path to the ultraplan JSON file to validate |

Flags: | Flag | Description | |------|-------------| | --json | Output results as JSON for CI/CD integration |

Validation checks: - Valid JSON syntax - Required fields present - Task dependency validity (no cycles, no missing references) - File conflict detection between parallel tasks - Warnings for high complexity tasks

Examples:

# Validate a plan file
claudio validate .claudio-plan.json

# JSON output for CI/CD
claudio validate --json my-plan.json

# Validate before executing
claudio validate plan.json && claudio ultraplan --plan plan.json


claudio logs

View and filter session logs.

claudio logs [flags]

Flags: | Flag | Short | Description | |------|-------|-------------| | --session | -s | Session ID (default: most recent) | | --tail | -n | Number of lines to show, 0 for all (default: 50) | | --follow | -f | Follow log output in real-time | | --level | | Minimum level to show (debug/info/warn/error) | | --since | | Show logs since duration (e.g., 1h, 30m) | | --grep | | Filter by regex pattern |

Examples:

# Show last 50 lines
claudio logs

# Show all logs
claudio logs -n 0

# Follow logs in real-time
claudio logs -f

# Filter by level and pattern
claudio logs --level warn --grep "conflict"

# Logs from specific session
claudio logs -s abc123 --since 1h


claudio completion

Generate shell autocompletion scripts.

claudio completion [shell]

Supported shells: - bash - zsh - fish - powershell

Examples:

# Bash
claudio completion bash > /etc/bash_completion.d/claudio

# Zsh
claudio completion zsh > "${fpath[1]}/_claudio"

# Fish
claudio completion fish > ~/.config/fish/completions/claudio.fish


Exit Codes

Code Meaning
0 Success
1 General error
2 Invalid arguments
130 Interrupted (Ctrl+C)

Environment Variables

All configuration options can be set via environment variables:

CLAUDIO_COMPLETION_DEFAULT_ACTION=auto_pr
CLAUDIO_BRANCH_PREFIX=feature
CLAUDIO_PR_DRAFT=true

See Configuration Reference for all options.