Commands

Core Commands

codemie --help                   # Show all commands and options
codemie --version                # Show version information
codemie --task "task"            # Execute single task with built-in agent and exit

codemie setup                    # Interactive configuration wizard
codemie profile <command>        # Manage provider profiles
codemie analytics [options]      # View usage analytics
codemie log [options]            # View and manage debug logs and sessions
codemie workflow <command>       # Manage CI/CD workflows
codemie list [options]           # List all available agents
codemie install [agent]          # Install an agent
codemie uninstall [agent]        # Uninstall an agent
codemie update [agent]           # Update installed agents
codemie self-update              # Update CodeMie CLI itself
codemie doctor [options]         # Health check and diagnostics
codemie plugin <command>         # Manage native plugins
codemie mcp-proxy <url>          # Stdio-to-HTTP MCP proxy with OAuth support
codemie version                  # Show version information

Global Options

--task <task>            # Execute a single task using the built-in agent and exit
-s, --silent             # Enable silent mode
--help                   # Display help for command
--version                # Output the version number

Agent Shortcuts

Direct access to agents with automatic configuration.

Common Options (All Agents)

All agent shortcuts support these options:

--help                   # Display help for agent
--version                # Show agent version
--profile <name>         # Use specific provider profile
--provider <provider>    # Override provider (ai-run-sso, litellm, ollama)
-m, --model <model>      # Override model
--api-key <key>          # Override API key
--base-url <url>         # Override base URL
--timeout <seconds>      # Override timeout (in seconds)
-s, --silent             # Enable silent mode
--task <prompt>          # Run a single task in headless (non-interactive) mode and exit
--jwt-token <token>      # JWT token for authentication (overrides config and CODEMIE_JWT_TOKEN)

Built-in Agent (codemie-code)

codemie-code                     # Interactive mode
codemie-code "message"           # Start with initial message
codemie-code health              # Health check
codemie-code --help              # Show help with all options

# With configuration overrides
codemie-code --profile work-litellm "analyze codebase"
codemie-code --model claude-sonnet-4-6 "review code"
codemie-code --provider ollama --model codellama "generate tests"

External Agents

All external agents share the same command pattern:

# Basic usage
codemie-claude "message"         # Claude Code agent
codemie-claude-acp               # Claude Code ACP (invoked by IDEs)
codemie-gemini "message"         # Gemini CLI agent

# Health checks
codemie-claude health
codemie-gemini health

# Note: codemie-claude-acp doesn't have interactive mode or health check
# It's designed to be invoked by IDEs via ACP protocol

# With configuration overrides
codemie-claude --model claude-sonnet-4-6 --api-key sk-... "review code"
codemie-gemini -m gemini-2.5-flash --api-key key "optimize performance"
# With profile selection
codemie-claude --profile personal-openai "review PR"
codemie-gemini --profile google-direct "analyze code"

# Agent-specific options (pass-through to underlying CLI)
codemie-claude --context large -p "review code"      # -p = print mode (non-interactive)
codemie-gemini -p "your prompt"                      # -p for gemini's non-interactive mode

# Implement planned task without asking any questions (silent mode)
codemie-claude --task "Implement task 1" --silent --dangerously-skip-permissions --output-format stream-json --verbose

Note: Configuration options (--profile, --model, etc.) are handled by CodeMie CLI wrapper. All other options are passed directly to the underlying agent binary.

Headless Mode (`--task`)

The --task flag runs a single prompt non-interactively: the agent executes the task, prints the result, and exits. No user interaction is required. This is the primary way to use CodeMie agents in CI/CD pipelines and automated scripts.

How It Works

Each agent maps --task to its own non-interactive mechanism:

Agent	Underlying flag/command	Behaviour
`codemie-claude`	`-p <prompt>` (print mode)	Runs prompt, prints output, exits
`codemie-gemini`	`-p <prompt>`	Runs prompt, prints output, exits
`codemie-opencode`	`opencode run <prompt>`	Runs task via `run` subcommand, exits

Basic Usage

# Claude
codemie-claude --task "Summarize the recent changes in this repo"

# Gemini
codemie-gemini --task "Explain the architecture of this project"

# OpenCode
codemie-opencode --task "Review the code in src/ for security issues"

With Profile or Model Override

codemie-claude --profile work --task "Fix the failing tests"
codemie-gemini --model gemini-2.5-flash --task "Generate a changelog for this release"

Capturing Output

# Redirect stdout to a file
codemie-claude --task "Review this PR" > review.txt

# Capture both stdout and stderr
codemie-claude --task "Analyze codebase" > output.txt 2>&1

Non-Interactive Flags (Claude-specific)

When running headless Claude tasks, combine --task with these pass-through flags for full automation:

codemie-claude \
  --task "Implement the changes described in SPEC.md" \
  --silent \
  --dangerously-skip-permissions \
  --output-format stream-json \
  --verbose

Headless Mode with JWT Authentication

For CI/CD environments where SSO login is not possible, combine --task with --jwt-token:

# Single command — no prior setup required
codemie-claude \
  --jwt-token "$CI_JWT_TOKEN" \
  --task "Review the changes in this commit and list any issues"

# With base URL for environments not yet configured
codemie-claude \
  --jwt-token "$CI_JWT_TOKEN" \
  --base-url "https://codemie.lab.epam.com" \
  --task "Run a security review of the staged files"

See JWT Bearer Authorization for full token setup options.

CI/CD Examples

GitHub Actions:

- name: AI Code Review
  run: |
    codemie-claude \
      --jwt-token "${{ secrets.CODEMIE_JWT_TOKEN }}" \
      --task "Review the changes in this PR and report any issues" \
      --silent

GitLab CI:

ai-review:
  script:
    - codemie-claude
        --jwt-token "$CODEMIE_JWT_TOKEN"
        --task "Review changes in this commit"
        --silent

Shell script:

#!/bin/bash
RESULT=$(codemie-claude --jwt-token "$TOKEN" --task "Analyze src/ for bugs" 2>&1)
echo "$RESULT" > /tmp/ai-review.txt

Profile Management Commands

Manage multiple provider configurations (work, personal, team, etc.) with separate profiles.

codemie profile                      # List all profiles with detailed information (default action)
codemie profile status               # Show active profile and authentication status
codemie profile switch <name>        # Switch to a different profile
codemie profile delete <name>        # Delete a profile
codemie profile rename <old> <new>   # Rename a profile
codemie profile login [--url <url>]  # Authenticate with AI/Run CodeMie SSO
codemie profile logout               # Clear SSO credentials
codemie profile refresh              # Refresh SSO credentials

Note: To create or update profiles, use codemie setup which provides an interactive wizard.

Profile List Details: The codemie profile command displays comprehensive information for each profile:

Profile name and active status
Provider (ai-run-sso, openai, azure, bedrock, litellm, gemini)
Base URL
Model
Timeout settings
Debug mode status
Masked API keys (for security)
Additional provider-specific settings

SSO Authentication: For profiles using AI/Run CodeMie SSO provider:

login - Opens browser for SSO authentication, stores credentials securely
logout - Clears stored SSO credentials
status - Shows active profile with auth status, prompts for re-auth if invalid
refresh - Re-authenticates with existing SSO configuration

Analytics Commands

Track and analyze your AI agent usage across all agents.

# View analytics summary
codemie analytics                # Show all analytics with aggregated metrics

# Filter by criteria
codemie analytics --project codemie-code        # Filter by project
codemie analytics --agent claude                # Filter by agent
codemie analytics --branch main                 # Filter by branch
codemie analytics --from 2025-12-01             # Date range filter
codemie analytics --last 7d                     # Last 7 days

# Output options
codemie analytics --verbose                     # Detailed session breakdown
codemie analytics --export json                 # Export to JSON
codemie analytics --export csv -o report.csv    # Export to CSV

# View specific session
codemie analytics --session abc-123-def         # Single session details

Analytics Features:

Hierarchical aggregation: Root → Projects → Branches → Sessions
Session metrics: Duration, turns, tokens, costs
Model distribution across all sessions
Tool usage breakdown with success/failure rates
Language/format statistics (lines added, files created/modified)
Cache hit rates and token efficiency metrics
Export to JSON/CSV for external analysis
Privacy-first (local storage at ~/.codemie/metrics/)

Example Workflows:

# Weekly summary
codemie analytics --last 7d

# Project-specific with details
codemie analytics --project my-project --verbose

# Cost tracking
codemie analytics --from 2025-12-01 --to 2025-12-07 --export csv -o weekly-costs.csv

# Agent comparison
codemie analytics --agent claude
codemie analytics --agent gemini

Log Management Commands

View, filter, and manage debug logs and agent sessions.

# View recent logs
codemie log                             # Show last 50 lines
codemie log -n 100                      # Show last 100 lines
codemie log -v                          # Verbose mode with session IDs

# Filter logs
codemie log --session abc-123           # Filter by session ID
codemie log --agent claude              # Filter by agent
codemie log --level error               # Show only errors
codemie log --profile work              # Filter by profile

# Date filtering
codemie log --from 2026-02-01           # From specific date
codemie log --to 2026-02-04             # Until specific date
codemie log --last 7d                   # Last 7 days
codemie log --last 24h                  # Last 24 hours
codemie log --last 30m                  # Last 30 minutes

# Pattern search
codemie log --grep "error"              # Search for pattern
codemie log --grep "sync" --last 1d     # Search in recent logs

# Session management
codemie log session <id>                # View specific session details
codemie log session <id> -v             # Include conversation history
codemie log list-sessions               # List all sessions
codemie log list-sessions --agent claude --last 7d

# Real-time monitoring
codemie log follow                      # Follow logs in real-time (tail -f)
codemie log follow --level error        # Follow only errors
codemie log follow --agent claude       # Follow specific agent

# Cleanup
codemie log clean --dry-run             # Preview cleanup
codemie log clean --days 10             # Keep last 10 days
codemie log clean --days 30 --sessions  # Also clean sessions
codemie log clean --yes                 # Skip confirmation

# Export logs
codemie log --format json -o logs.json          # Export to JSON
codemie log --format jsonl -o logs.jsonl        # Export to JSONL
codemie log --last 7d --format json -o week.json

Log Features:

Real-time log viewing with colorized output
Multiple filtering options (session, agent, level, date, pattern)
Session inspection with conversation history
Live log following (tail -f style)
Cleanup old logs and sessions
Export to JSON/JSONL for analysis
Graceful handling of missing/corrupted files
Local storage at ~/.codemie/logs/ and ~/.codemie/sessions/

Log Levels:

debug - Detailed debugging information
info - General informational messages
warn - Warning messages
error - Error messages

Example Workflows:

# Troubleshoot recent errors
codemie log --level error --last 1h

# Investigate specific session
codemie log --session abc-123-def -v

# Monitor agent activity
codemie log follow --agent claude --level info

# Weekly log analysis
codemie log --last 7d --format json -o weekly-logs.json

# Clean old logs (keep last 10 days)
codemie log clean --days 10 --dry-run
codemie log clean --days 10 --yes

# Search for specific issues
codemie log --grep "timeout" --last 24h

OpenCode Metrics Commands

Process OpenCode session data to extract metrics and sync to analytics system.

# Process specific session
codemie opencode-metrics --session <session-id>

# Discover and process all recent sessions
codemie opencode-metrics --discover

# Verbose output with detailed processing info
codemie opencode-metrics --discover --verbose

Options:

-s, --session <id> - Process specific OpenCode session by ID
-d, --discover - Discover and process all unprocessed sessions (last 30 days)
-v, --verbose - Show detailed processing output

Features:

Automatic session discovery from OpenCode storage
Token usage extraction (input, output, total)
Cost calculation based on model pricing
Session duration tracking
Conversation extraction
JSONL delta generation for sync
Deduplication (skips recently processed sessions)

Session Storage Locations:

Linux: ~/.local/share/opencode/storage/
macOS: ~/Library/Application Support/opencode/storage/
Windows: %LOCALAPPDATA%\opencode\storage\

Example Workflows:

# Process all recent OpenCode sessions
codemie opencode-metrics --discover --verbose

# Check specific session metrics
codemie opencode-metrics --session ses_abc123def456

# View results in analytics
codemie analytics --agent opencode

Note: Metrics are automatically extracted when OpenCode sessions end (via onSessionEnd lifecycle hook). Manual processing is useful for:

Retroactive processing of old sessions
Troubleshooting sync issues
Verifying metrics extraction

Workflow Commands

Install CI/CD workflows for automated code review and generation.

# List available workflows
codemie workflow list                    # All workflows
codemie workflow list --installed        # Only installed

# Install workflows
codemie workflow install pr-review       # PR review workflow
codemie workflow install inline-fix      # Quick fixes from comments
codemie workflow install code-ci         # Full feature implementation
codemie workflow install --interactive   # Interactive installation

# Uninstall workflows
codemie workflow uninstall pr-review     # Remove workflow

Available Workflows:

pr-review - Automated code review on pull requests
inline-fix - Quick code fixes from PR comments
code-ci - Full feature implementation from issues

Supported Platforms:

GitHub Actions (auto-detected from .git/config)
GitLab CI (auto-detected from .git/config)

Plugin Commands

Manage native plugins (Anthropic format) for extending CodeMie Code with reusable packages of skills, commands, agents, hooks, and MCP servers.

# List all discovered plugins
codemie plugin list [--cwd <path>]

# Install a plugin from a local path
codemie plugin install <path>

# Remove a plugin from the cache
codemie plugin uninstall <name>

# Enable a disabled plugin
codemie plugin enable <name>

# Disable a plugin without removing it
codemie plugin disable <name>

Plugin Sources (priority order):

CLI flag --plugin-dir (highest)
Project .codemie/plugins/
User cache ~/.codemie/plugins/cache/
Config plugins.dirs (lowest)

For full documentation, see Plugin System.

MCP Proxy Command

Run a stdio-to-HTTP bridge that connects MCP clients (like Claude Code) to remote MCP servers, handling OAuth 2.0 authorization automatically.

codemie mcp-proxy <url>

Arguments:

<url> — Remote MCP server URL (must be a valid HTTP/HTTPS URL)

Features:

Stdio-to-HTTP bridge (JSON-RPC over stdio ↔ streamable HTTP transport)
Automatic OAuth 2.0 with Dynamic Client Registration
Per-origin cookie jar for session persistence
Browser-based authorization with ephemeral localhost callback server
Graceful shutdown on SIGINT/SIGTERM

Registering with Claude Code

Use claude mcp add to register the proxy as an MCP server:

# Using the global binary (requires global install)
claude mcp add my-server -- codemie-mcp-proxy "https://mcp-server.example.com/sse"

# Using node directly (works without global install)
claude mcp add my-server -- node /path/to/bin/codemie-mcp-proxy.js "https://mcp-server.example.com/sse"

Or configure .mcp.json directly:

{
  "mcpServers": {
    "my-server": {
      "type": "stdio",
      "command": "codemie-mcp-proxy",
      "args": ["https://mcp-server.example.com/sse"],
      "env": {
        "MCP_CLIENT_NAME": "Claude Code (my-server)"
      }
    }
  }
}

Environment Variables

Variable	Default	Description
`MCP_CLIENT_NAME`	`CodeMie CLI`	Client name used in OAuth Dynamic Client Registration
`MCP_PROXY_DEBUG`	(unset)	Set to `true` to enable verbose proxy logging
`CODEMIE_DEBUG`	(unset)	Set to `true` to enable general debug logging

OAuth Flow

When the remote MCP server returns 401 Unauthorized:

Discover resource metadata and authorization server metadata
Register a client dynamically (client_name from MCP_CLIENT_NAME)
Open the user's browser for authorization
Receive the authorization code via ephemeral localhost callback server
Exchange the code for tokens
Retry the original request with the Bearer token

All state is in-memory only — re-authorization is required each session.

Logs

Proxy logs are written to ~/.codemie/logs/mcp-proxy.log. Enable verbose logging with MCP_PROXY_DEBUG=true.

Troubleshooting

Issue	Cause	Solution
`Invalid MCP server URL`	URL argument is malformed	Verify the URL is a valid HTTP/HTTPS URL
Browser doesn't open	System `open`/`xdg-open` not available	Open the URL printed in logs manually
OAuth timeout	User didn't complete browser auth in 2 minutes	Re-run the command and complete auth faster
`ECONNREFUSED`	Remote MCP server is unreachable	Check the URL and network connectivity
No tools appearing	OAuth flow not completed	Check `~/.codemie/logs/mcp-proxy.log` for errors

Architecture

For implementation details including the SSO proxy plugin, URL rewriting, and SSRF protection, see Proxy Architecture.

Detailed Command Reference

`codemie setup`

Interactive configuration wizard for setting up AI providers.

Usage:

codemie setup [options]

Features:

Multi-provider support (AI-Run SSO, OpenAI, Azure, Bedrock, LiteLLM, Ollama)
Real-time model fetching and validation
Health endpoint testing during setup
Profile management (add new or update existing)
Credential validation before saving

`codemie list`

List all available AI coding agents.

Usage:

codemie list [options]

Options:

-i, --installed - Show only installed agents

Output:

Agent name and display name
Installation status
Version (if installed)
Description

`codemie install [agent]`

Install an external AI coding agent.

Usage:

codemie install <agent>

Supported Agents:

claude - Claude Code (npm-based)
claude-acp - Claude Code ACP adapter for IDE integration (npm-based)
gemini - Gemini CLI (npm-based)
opencode - OpenCode AI assistant (npm-based)

`codemie uninstall [agent]`

Uninstall an external AI coding agent.

Usage:

codemie uninstall <agent>

`codemie update [agent]`

Update installed AI coding agents to their latest versions.

Usage:

# Update specific agent
codemie update <agent>

# Check for updates without installing
codemie update <agent> --check

# Interactive update (checks all agents)
codemie update

# Check all agents for updates
codemie update --check

Options:

-c, --check - Check for updates without installing

Features:

Checks npm registry for latest versions
Supports interactive multi-agent selection
Shows current vs. latest version comparison
Special handling for Claude Code (uses verified versions)
Uses --force flag to handle directory conflicts during updates

Examples:

# Update Claude Code to latest verified version
codemie update claude

# Check if Gemini has updates
codemie update gemini --check

# Interactive: select which agents to update
codemie update

Note: This command updates external agents (Claude Code, Gemini, etc.). To update the CodeMie CLI itself, use codemie self-update.

`codemie self-update`

Update CodeMie CLI to the latest version from npm.

Usage:

# Update CodeMie CLI
codemie self-update

# Check for updates without installing
codemie self-update --check

Options:

-c, --check - Check for updates without installing

Features:

Fast version check with 5-second timeout
Automatic update on startup (configurable via CODEMIE_AUTO_UPDATE)
Uses --force flag to handle directory conflicts
Shows current vs. latest version comparison

Auto-Update Behavior:

By default, CodeMie CLI automatically checks for updates on startup with smart rate limiting:

# Default: Silent auto-update (no user interaction)
codemie --version
# First run: Checks for updates (5s max)
# Subsequent runs within 24h: Instant (skips check)

# Prompt before updating
export CODEMIE_AUTO_UPDATE=false
codemie --version

# Explicit silent auto-update
export CODEMIE_AUTO_UPDATE=true
codemie --version

Performance & Rate Limiting:

Update checks are rate-limited to once per 24 hours by default
First invocation may take up to 5 seconds (network check)
Subsequent invocations within the interval are instant (no network call)
Prevents blocking on every CLI startup
Cache stored in ~/.codemie/.last-update-check

Environment Variables:

CODEMIE_AUTO_UPDATE=true (default) - Silently auto-update in background
CODEMIE_AUTO_UPDATE=false - Show update prompt and ask for confirmation
CODEMIE_UPDATE_CHECK_INTERVAL - Time between checks in ms (default: 86400000 = 24h)

Examples:

# Check for CLI updates
codemie self-update --check

# Update CLI immediately
codemie self-update

# Disable auto-update (add to ~/.bashrc or ~/.zshrc)
export CODEMIE_AUTO_UPDATE=false

Note: Auto-update checks are non-blocking and won't prevent CLI from starting if they fail. The update takes effect on the next command execution.

`codemie doctor`

Check system health and configuration.

Usage:

codemie doctor [options]

Options:

-v, --verbose - Enable verbose debug output with detailed API logs

Checks:

Node.js version (requires >=20.0.0)
Python version (if using Python-based agents)
Git installation and configuration
AWS CLI (if using Bedrock)
Installed agents and their versions
Provider connectivity and health endpoints
Configuration file validity

`codemie profile`

Manage multiple provider configurations and SSO authentication.

Usage:

codemie profile                         # List all profiles with details (default action)
codemie profile status                  # Show active profile and authentication status
codemie profile switch [profile]        # Switch active profile
codemie profile delete [profile]        # Delete a profile
codemie profile rename <old> <new>      # Rename a profile
codemie profile login [--url <url>]     # Authenticate with AI/Run CodeMie SSO
codemie profile logout                  # Clear SSO credentials and logout
codemie profile refresh                 # Refresh SSO credentials

Profile Management:

Active profile indicator (●)
Profile name
Provider type
Model configuration
Base URL
Masked API key (for security)
Timeout and other settings

SSO Authentication:

login - Opens browser for SSO authentication, stores credentials securely
logout - Clears stored SSO credentials
status - Shows active profile with auth status, prompts for re-auth if invalid
refresh - Re-authenticates with existing SSO configuration

`codemie workflow`

Manage CI/CD workflow templates for GitHub Actions and GitLab CI.

Subcommands:

codemie workflow list [options]                     # List available workflow templates
codemie workflow install [options] <workflow-id>    # Install a workflow template
codemie workflow uninstall [options] <workflow-id>  # Uninstall a workflow

List Options:

--installed - Show only installed workflows

Install Options:

-i, --interactive - Interactive mode with helpful prompts
--timeout <minutes> - Workflow timeout (default: 15)
--max-turns <number> - Maximum AI conversation turns (default: 50)
--environment <env> - GitHub environment for protection rules

Available Workflows:

pr-review - Automated code review on pull requests
inline-fix - Quick fixes from PR comments mentioning @codemie
code-ci - Full feature implementation from issues

`codemie analytics`

Display aggregated metrics and analytics from agent usage sessions.

Usage:

codemie analytics [options]

Filter Options:

--session <id> - Filter by session ID
--project <pattern> - Filter by project path (basename, partial, or full path)
--agent <name> - Filter by agent name (claude, gemini, etc.)
--branch <name> - Filter by git branch
--from <date> - Filter sessions from date (YYYY-MM-DD)
--to <date> - Filter sessions to date (YYYY-MM-DD)
--last <duration> - Filter sessions from last duration (e.g., 7d, 24h)

Output Options:

-v, --verbose - Show detailed session-level breakdown
--export <format> - Export to file (json or csv)
-o, --output <path> - Output file path (default: ./codemie-analytics-YYYY-MM-DD.{format})

Metrics Displayed:

Session count and duration
Token usage (input/output/total)
Cost estimates
Model distribution
Tool usage statistics
Cache hit rates
Language/format statistics

For detailed usage examples and filtering options, see the Analytics Commands section above.

`codemie log`

View, filter, and manage debug logs and agent sessions.

Usage:

codemie log [options]              # View recent debug logs
codemie log <subcommand> [options] # Execute subcommand

Main Command Options:

--session <id> - Filter by session ID
--agent <name> - Filter by agent (claude, gemini, etc.)
--profile <name> - Filter by profile name
--level <level> - Filter by log level (debug, info, warn, error)
--from <date> - Filter from date (YYYY-MM-DD)
--to <date> - Filter to date (YYYY-MM-DD)
--last <duration> - Filter last duration (e.g., 7d, 24h, 30m)
--grep <pattern> - Search pattern (supports regex)
-n, --lines <number> - Number of lines to show (default: 50)
-v, --verbose - Show full details including session IDs and profiles
--format <format> - Output format (text, json, jsonl)
--no-color - Disable color output
-o, --output <path> - Write to file instead of stdout

Subcommands:

1. codemie log debug [options]

View debug logs (alias for default behavior).

codemie log debug                  # Same as 'codemie log'
codemie log debug --level error    # Show only errors
codemie log debug --agent claude   # Claude logs only

2. codemie log session <id> [options]

View specific session details.

codemie log session abc-123-def-456        # Basic session info
codemie log session abc-123-def-456 -v     # Include conversation
codemie log session abc-123 --format json  # JSON output

Options:

-v, --verbose - Show conversation details
--format <format> - Output format (text, json)
--no-color - Disable color output

3. codemie log list-sessions [options]

List all sessions with filtering and sorting.

codemie log list-sessions                  # All sessions
codemie log list-sessions --agent claude   # Claude sessions only
codemie log list-sessions --last 7d        # Last week
codemie log list-sessions --sort duration  # Sort by duration

Options:

--agent <name> - Filter by agent
--from <date> - Filter from date
--to <date> - Filter to date
--last <duration> - Filter last duration
--sort <field> - Sort by field (time, duration, agent)
--reverse - Reverse sort order
--format <format> - Output format (text, json)
--no-color - Disable color output

4. codemie log follow [options]

Follow logs in real-time (tail -f style).

codemie log follow                    # Follow all logs
codemie log follow --level error      # Follow errors only
codemie log follow --agent claude     # Follow Claude agent
codemie log follow --grep "timeout"   # Follow matching pattern

Options:

--agent <name> - Filter by agent
--level <level> - Filter by log level
--grep <pattern> - Search pattern
-v, --verbose - Show full details
--no-color - Disable color output

Press Ctrl+C to stop following.

5. codemie log clean [options]

Clean up old logs and sessions.

codemie log clean --dry-run              # Preview what would be deleted
codemie log clean --days 10              # Keep last 10 days
codemie log clean --days 30 --sessions   # Also delete sessions
codemie log clean --yes                  # Skip confirmation

Options:

--days <number> - Retention period in days (default: 5)
--sessions - Also delete old sessions (not just debug logs)
--dry-run - Preview without deleting
--yes - Skip confirmation prompt

Log Storage:

Debug logs: ~/.codemie/logs/debug-YYYY-MM-DD.log
Session data: ~/.codemie/sessions/
Automatic daily rotation
Default retention: 5 days for logs, unlimited for sessions

Log Format:

Each log entry contains:

Timestamp (ISO 8601)
Level (DEBUG, INFO, WARN, ERROR)
Agent name
Session ID
Profile (optional)
Message

Example:

[2026-02-04T10:30:45.123Z] [INFO] [claude] [abc-123] [work] Session started

Examples:

# Quick troubleshooting
codemie log --level error --last 1h

# Detailed session investigation
codemie log session abc-123-def-456 -v

# Monitor specific agent
codemie log follow --agent claude --level info

# Export recent logs for analysis
codemie log --last 7d --format json -o weekly.json

# Clean old data (preview first)
codemie log clean --days 10 --dry-run
codemie log clean --days 10 --yes

# Search for specific issues
codemie log --grep "connection refused" --last 24h

# Verbose output with all context
codemie log -v -n 100 --last 1d

# List recent sessions sorted by duration
codemie log list-sessions --last 7d --sort duration --reverse

Tips:

Use --dry-run before cleaning to preview deletions
Combine filters for precise searches (--agent --level --last)
Export to JSON for programmatic analysis or bug reports
Use follow mode for real-time monitoring during development
Session data is never auto-deleted (only via explicit clean --sessions)

`codemie version`

Show version information for CodeMie CLI.

Usage:

codemie version

Output:

CLI version
Node.js version
Package name and description

FilesExpand file tree

COMMANDS.md

Latest commit

History

COMMANDS.md

File metadata and controls

Commands

Core Commands

Global Options

Agent Shortcuts

Common Options (All Agents)

Built-in Agent (codemie-code)

External Agents

Headless Mode (--task)

How It Works

Basic Usage

With Profile or Model Override

Capturing Output

Non-Interactive Flags (Claude-specific)

Headless Mode with JWT Authentication

CI/CD Examples

Profile Management Commands

Analytics Commands

Log Management Commands

OpenCode Metrics Commands

Workflow Commands

Plugin Commands

MCP Proxy Command

Registering with Claude Code

Environment Variables

OAuth Flow

Logs

Troubleshooting

Architecture

Detailed Command Reference

codemie setup

codemie list

codemie install [agent]

codemie uninstall [agent]

codemie update [agent]

codemie self-update

codemie doctor

codemie profile

codemie workflow

codemie analytics

codemie log

codemie version

Headless Mode (`--task`)

`codemie setup`

`codemie list`

`codemie install [agent]`

`codemie uninstall [agent]`

`codemie update [agent]`

`codemie self-update`

`codemie doctor`

`codemie profile`

`codemie workflow`

`codemie analytics`

`codemie log`

`codemie version`