AI/Run CodeMie CLI

Unified AI Coding Assistant CLI - Manage Claude Code, OpenAI Codex, GitHub Copilot CLI, Google Gemini, OpenCode, Pi, Kimi Code, and custom AI agents from one powerful command-line interface. Multi-provider support (CodeMie SSO, Bearer Auth, LiteLLM, AWS Bedrock, Ollama, Anthropic Subscription, Moonshot Subscription). Built-in native agent with file operations, command execution, planning mode, and plugins. Cross-platform support for Windows, Linux, and macOS.

---

---

Why CodeMie CLI?

CodeMie CLI is the all-in-one AI coding assistant for developers.

• ✨ One CLI, Multiple AI Agents - Switch between Claude Code, OpenAI Codex, GitHub Copilot CLI, Gemini, OpenCode, Pi, Kimi Code, and built-in agent.
• 🔄 Multi-Provider Support - CodeMie SSO, Bearer Authorization, LiteLLM, AWS Bedrock, Ollama, Anthropic Subscription, and Moonshot Subscription.
• 🚀 Built-in Agent - codemie-code ships with the CLI: file operations, command execution, planning mode, and native plugins.
• 🖥️ Cross-Platform - Full support for Windows, Linux, and macOS with platform-specific optimizations.
• 🔗 MCP Proxy - Connect to remote MCP servers with automatic OAuth authorization.
• 🔐 Enterprise Ready - SSO and JWT authentication, audit logging, and role-based access.
• ⚡ Productivity Boost - Code review, refactoring, test generation, and bug fixing.
• 🎯 Profile Management - Manage work, personal, and team configurations separately.
• 🧩 CodeMie Assistants in Claude - Connect your available CodeMie assistants as Claude subagents or skills.
• 🛠️ CodeMie Platform Skills - Install CodeMie platform skills directly as Claude Code slash commands with auto-sync.
• 📊 Usage Analytics - Track and analyze AI usage across all agents with detailed insights.
• 🔧 CI/CD Workflows - Automated code review, fixes, and feature implementation.

Perfect for developers seeking a powerful alternative to GitHub Copilot or Cursor.

Quick Start

Install CodeMie using the instructions for your shell, then run:

codemie setup
codemie doctor
codemie install claude --supported
codemie install codex --supported
codemie install copilot
codemie-claude "Review my API code"
codemie-codex "Refactor this service"
codemie-copilot --task "Summarize this module"
codemie --task "Generate unit tests"
codemie skills find pdf                    # discover agent skills (EPAM internal + skills.sh)
claude mcp add my-server -- codemie-mcp-proxy "https://mcp-server.example.com/sse"

Prefer not to install globally? Use npx with the full package name:

npx @codemieai/code setup
npx @codemieai/code doctor
npx @codemieai/code install claude --supported
# Note: Agent shortcuts require global installation

Installation

Native Bootstrap Installers

For Windows and macOS, CodeMie ships two installer options:

• GUI installers — a signed .dmg (macOS) and a .exe wizard (Windows) that guide you through installation with no terminal required. Download CodeMie Connect 2.0.1 (macOS aarch64) or browse the macOS install folder / Windows install folder and run the file.
• Script installers — plain shell/PowerShell scripts stored in this repo that install via npm. Prefer these for CI, headless machines, or when the GUI installer is unavailable.

The script installers are plain scripts stored in this public GitHub repo, so they do not require a Windows-built .exe or a private Artifactory mirror.

The bootstrap path is recommended for non-technical users and managed enterprise machines because it:

• avoids PowerShell npm.ps1 execution-policy failures on Windows,
• avoids global npm permission errors such as macOS EACCES,
• installs into a user-writable location where possible,
• checks Node.js, npm, registry access, and CodeMie package visibility before installing,
• prints actionable remediation when the enterprise npm registry is not configured correctly.

The examples below use GitHub raw URLs from the main branch. For reproducible installs, replace main with a release tag such as v0.0.57. Enterprise teams can mirror the same scripts to Artifactory later by setting CODEMIE_INSTALL_URL to the mirrored script directory.

Channel selection is not implemented in the bootstrap scripts yet. To pin a version on Windows PowerShell, pass -Version 0.0.57. To pin a version on macOS, Linux, or WSL, set CODEMIE_PACKAGE_VERSION=0.0.57 before running the install command.

Windows PowerShell

The Windows bootstrapper installs CodeMie in user-local portable mode by default and calls npm.cmd directly, so it does not permanently change PowerShell execution policy.

irm https://raw.githubusercontent.com/codemie-ai/codemie-code/main/install/windows/install.ps1 | iex

To pass explicit options:

& ([scriptblock]::Create((irm https://raw.githubusercontent.com/codemie-ai/codemie-code/main/install/windows/install.ps1))) -RegistryUrl https://registry.npmjs.org/

Windows CMD

Use this fallback when PowerShell copy-paste guidance is not practical:

curl -fsSL https://raw.githubusercontent.com/codemie-ai/codemie-code/main/install/windows/install.cmd -o install.cmd && install.cmd && del install.cmd

macOS

The macOS bootstrapper uses npm global installation only when it is user-writable. If global npm is not writable, it configures a user-local npm prefix instead.

curl -fsSL https://raw.githubusercontent.com/codemie-ai/codemie-code/main/install/macos/install.sh | bash

To install a specific package version:

curl -fsSL https://raw.githubusercontent.com/codemie-ai/codemie-code/main/install/macos/install.sh | env CODEMIE_PACKAGE_VERSION=0.0.57 bash

Linux and WSL

Use the same shell bootstrapper for Linux and WSL:

curl -fsSL https://raw.githubusercontent.com/codemie-ai/codemie-code/main/install/macos/install.sh | bash

npm Fallback

Use npm fallback only when Node.js 20+ and npm global installs are already configured correctly:

npm install -g @codemieai/code
codemie --help

Local/Project npm Installation

For project-specific usage:

npm install @codemieai/code

# Use with npx
npx @codemieai/code --help

Note: Agent shortcuts (codemie-claude, codemie-codex, codemie-code, codemie-opencode, codemie-pi, codemie-kimi, etc.) require global installation.

Installation Troubleshooting

If PowerShell reports that npm.ps1 cannot be loaded, use the CodeMie bootstrap installer. It calls npm.cmd directly and does not permanently change your execution policy.

If npm reports EACCES on macOS, use the bootstrap installer or configure npm to use a user-local prefix.

If npm reports 404 Not Found, verify that you are installing @codemieai/code, not codemie, and that your enterprise npm virtual registry exposes the @codemieai scope.

If the installer says @codemieai/code is not visible in the registry, ask IT to expose the package through the approved virtual npm repository.

From Source

git clone https://github.com/codemie-ai/codemie-code.git
cd codemie-code
npm install
npm run build && npm link

Verify Installation

codemie --help
codemie doctor

Usage

The CodeMie CLI provides two ways to interact with AI agents:

Built-in Agent (CodeMie Native)

codemie-code ships with the CLI — no separate install. It is a native CodeMie binary (@codemieai/codemie-opencode, platform-specific), pre-wired to the CodeMie proxy, with file/shell tools, planning mode, and native plugin support.

codemie-code                                  # interactive session
codemie-code "Help me refactor this component"
codemie-code --task "Generate unit tests"     # run once and exit
codemie-code --plan                           # planning mode
codemie-code --plan-only                      # plan without executing
codemie-code --plugin-dir ./my-plugins        # load native plugins
codemie-code --debug                          # debug logging

Providers: CodeMie SSO, Bearer Auth, LiteLLM, AWS Bedrock, Ollama.

External Agents

CodeMie installs external agents, routes them through the CodeMie proxy, and tracks their sessions.

Agent	Install	Shortcut	Upstream package
Claude Code	codemie install claude --supported	codemie-claude	@anthropic-ai/claude-code
Claude Code ACP	codemie install claude-acp	codemie-claude-acp	@zed-industries/claude-code-acp
OpenAI Codex	codemie install codex --supported	codemie-codex	@openai/codex
Gemini CLI	codemie install gemini	codemie-gemini	@google/gemini-cli
OpenCode	codemie install opencode	codemie-opencode	opencode-ai
Pi	codemie install pi	codemie-pi	@earendil-works/pi-coding-agent
Kimi Code	codemie install kimi	codemie-kimi	@moonshot-ai/kimi-code
Kimi Code ACP	codemie install kimi-acp	codemie-kimi-acp	@moonshot-ai/kimi-code (acp mode)
GitHub Copilot CLI	codemie install copilot	codemie-copilot	@github/copilot

ACP agents are launched by your IDE, not directly — see ACP Agent usage below.

codemie install claude --supported             # install
codemie install copilot                        # install GitHub Copilot CLI
codemie-claude "Review my API code"            # one-shot task
codemie-claude                                 # interactive session
codemie-copilot --task "Refactor this auth flow"
codemie uninstall copilot                      # uninstall GitHub Copilot CLI
codemie-codex --task "Refactor this auth flow" # same shape for every agent

GitHub Copilot CLI via CodeMie

CodeMie can install and launch GitHub Copilot CLI as a managed agent while keeping the internal registry identity copilot-cli.

Supported managed path for this release:

• Install: codemie install copilot
• Uninstall: codemie uninstall copilot
• Launch: codemie-copilot
• One-shot task: codemie-copilot --task "Explain this service"
• Supported providers: CodeMie SSO and LiteLLM

Requirements and behavior:

• Use an authenticated CodeMie profile (codemie setup).
• Managed Copilot runs are routed through CodeMie's provider/proxy path.
• The supported managed path does not require GitHub login, a GitHub Copilot subscription, or a GitHub personal access token.
• CodeMie blocks fallback to ambient GitHub credentials for this managed mode.
• Copilot conversation sync is attributed under copilot-cli in CodeMie.

ACP Agent usage in IDEs and Editors

Zed (~/.config/zed/settings.json):

{
  "agent_servers": {
    "claude": {
      "command": "codemie-claude-acp",
      "args": ["--profile", "work"]
    }
  }
}

IntelliJ IDEA (~/.jetbrains/acp.json):

{
  "default_mcp_settings": {},
  "agent_servers": {
    "Claude Code via CodeMie": {
      "command": "codemie-claude-acp"
    }
  }
}

Emacs (with acp.el):

(setq acp-claude-command "codemie-claude-acp")
(setq acp-claude-args '("--profile" "work"))

Version Management:

CodeMie manages agent versions to ensure compatibility:

codemie install claude --supported   # latest supported version (recommended)
codemie install claude 2.1.22        # pin a specific version
codemie install claude               # latest available version

Auto-updates are automatically disabled to maintain version control. CodeMie notifies you when running a different version than supported.

For more detailed information on the available agents, see the Agents Documentation.

Providers

A profile binds an agent to a provider. Run codemie setup to create one, or codemie profile to manage several (work, personal, team).

Provider	Auth	Use case
CodeMie SSO	enterprise SSO	Enterprise default — centralized model management, proxy routing, analytics
Bearer Authorization	JWT via CLI or env var	CI, service accounts, self-hosted gateways
LiteLLM	API key	Universal gateway to 100+ LLM providers (OpenAI, Azure, Vertex, …)
AWS Bedrock	AWS access key + secret	Claude, Llama, Mistral & more via Amazon Bedrock
Ollama	none	Local open-source models, offline work
Anthropic Subscription	native Claude Code login	Bring your own Claude subscription
Moonshot Subscription	native Kimi Code login	Bring your own Moonshot subscription (Kimi Code)

See Authentication and Configuration for setup details.

CodeMie Assistants as Claude Skills or Subagents

CodeMie can connect assistants available in your CodeMie account directly into Claude Code. Register them as Claude subagents and call them with @slug, or register them as Claude skills and invoke them with /slug.

# Pick assistants from your CodeMie account and choose how to register them
codemie setup assistants

During setup, choose:

• Claude Subagents - register selected assistants as @slug
• Claude Skills - register selected assistants as /slug
• Manual Configuration - choose skill or subagent per assistant

After registration, use them from Claude Code:

@api-reviewer Review this authentication flow
/release-checklist prepare a release checklist for this branch

You can also message a registered assistant directly through CodeMie:

codemie assistants chat "assistant-id" "Review this API design"

CodeMie Platform Skills in Claude

In addition to assistants, CodeMie platform skills can be installed directly as Claude Code slash commands.

# Browse and register CodeMie platform skills
codemie setup skills

During setup:

1. A disclaimer is shown — skills are installed without tools or MCP servers. If you need tools, create an assistant with the skill attached and use codemie setup assistants instead.
2. Choose storage scope: Global (available in all projects) or Local (project-scoped, overrides global).
3. Select which skills to register or unregister from your CodeMie account.

After registration, use them directly in Claude Code:

/skill-name run the skill

Skills are automatically synced on every Claude agent startup, so the local SKILL.md files stay up to date with the latest content from the CodeMie platform.

Tip: For skills that require MCP servers or tools, use codemie setup assistants instead.

Manage skills.sh and EPAM Skills (codemie skills)

codemie skills is a SSO-gated wrapper around the upstream skills.sh CLI. It lets you discover, install, update, and remove agent skills from any compatible catalog while keeping CodeMie's authentication, telemetry, and EPAM-internal catalog support in one place.

# Discover skills (two-section results: EPAM Internal first, public skills.sh second)
codemie skills find pdf
codemie skills find pdf --json
codemie skills find pdf --limit 25

# Install / update / remove skills via the upstream skills CLI
codemie skills add anthropics/skills --skill pdf --agent claude-code -y
codemie skills update                      # update everything in the current scope
codemie skills remove pdf -y               # remove a specific skill

# List installed skills (use --global for user-scope)
codemie skills list
codemie skills list --global --json

Notes:

• EPAM Internal catalog is opt-in. Until your team configures the internal endpoint, codemie skills find shows the friendly placeholder for the internal section and returns public results from skills.sh. Enable the internal catalog by exporting CODEMIE_SKILLS_SEARCH_URL or by adding skillsSearchUrl to your CodeMie profile (~/.codemie/codemie-cli.config.json).
• Authentication. Every codemie skills * subcommand requires an active CodeMie SSO session. Run codemie setup or codemie profile login first.
• Telemetry. A single lifecycle event is recorded per invocation (completed or failed). The raw query string is never sent.
• Pass-through. codemie skills find (no query) hands off to the upstream skills find interactive prompt, so the existing UX still works while the two-section view becomes the default for direct queries.

Claude Code Built-in Commands

When using Claude Code (codemie-claude), you get access to powerful built-in commands for project documentation:

Project Documentation:

# Generate AI-optimized docs (CLAUDE.md + guides). Can be added optional details after command as well
/codemie:codemie-init

# Generate project-specific subagents. Can be added optional details after command as well
/codemie:codemie-subagents

Memory Management:

# Capture important learnings
/memory-add

# Audit and update documentation
/memory-refresh

These commands analyze your actual codebase to create tailored documentation and specialized agents. See Claude Plugin Documentation for details.

OpenCode Session Metrics

When using OpenCode, CodeMie automatically extracts and tracks session metrics:

Manual Metrics Processing:

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

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

# Verbose output with details
codemie opencode-metrics --discover --verbose

Metrics are automatically extracted at session end and synced to the analytics system. Use codemie analytics to view comprehensive usage statistics across all agents.

Pi (codemie-pi)

Pi is an open-source coding agent harness (@earendil-works/pi-coding-agent).

codemie install pi                            # also installs the required Pi packages
codemie-pi "Add retry logic to the HTTP client"
codemie-pi                                    # interactive session
codemie-pi --continue                         # continue the latest session
codemie-pi --resume <session-id>              # open a specific session

What CodeMie configures for you:

What	How
Managed agent dir	~/.pi/agent is copied once to <project>/.pi/codemie/agent; Pi runs against the copy via PI_CODING_AGENT_DIR. Your skills, prompts, and extensions carry over — your Pi home is never modified.
Live model catalogue	models.json is regenerated each run with providers codemie-proxy (OpenAI-compatible) and codemie-anthropic (Claude). Model/provider come from your active profile.
Required packages	pi-mcp-adapter (Pi ships without built-in MCP), pi-subagents, superpowers — installed at setup.
Session analytics	An injected extension records tokens, tools, and models per session and syncs at session end; visible in codemie analytics.

Providers: CodeMie SSO, Bearer Auth, LiteLLM.

Claude Code Statusline

The CodeMie Statusline displays live budget usage, project name, git branch, model, context window percentage, and token counts at the bottom of every Claude Code session.

# Install (or update) the statusline
codemie install statusline

# Remove it
codemie uninstall statusline

Once installed, the statusline appears automatically in every claude session:

[my-project] $4.21/$50 (8%) | (main) | [claude-sonnet-4-5] | ctx:12% in:45.2k out:3.1k

The script is deployed to ~/.claude/codemie-budget-status.js and registered as a statusLine command in ~/.claude/settings.json. Budget values are cached for 60 seconds to avoid redundant API calls.

Commands

The CodeMie CLI has a rich set of commands for managing agents, configuration, and more.

codemie setup            # Interactive configuration wizard
codemie list             # List all available agents
codemie install <name>   # Install an agent or add-on (e.g. statusline)
codemie uninstall <name> # Remove an agent or add-on
codemie update <agent>   # Update installed agents
codemie self-update      # Update CodeMie CLI itself
codemie profile          # Manage provider profiles
codemie models list      # List models available for the current provider
codemie analytics        # View usage analytics (sessions, tokens, costs, tools)
codemie analytics --report --open   # Self-contained HTML dashboard (7 views, cost, date range, light/dark, no server)
codemie analytics --report --report-format json  # Same priced report data as JSON (--report-format html | json | both)
codemie workflow <cmd>   # Manage CI/CD workflows
codemie sdk <resource>   # Manage CodeMie platform assets (assistants, workflows, datasources, integrations, skills, users)
codemie skill <cmd>      # Manage skills for CodeMie agents (list, validate, reload)
codemie skills <cmd>     # Discover/install agent skills via skills.sh + EPAM catalog
codemie plugin <cmd>     # Manage native plugins (Anthropic format)
codemie mcp add          # Add an MCP server to an agent config
codemie mcp-proxy <url>  # Stdio-to-HTTP MCP proxy with OAuth
codemie proxy <cmd>      # Manage the local CodeMie proxy and its integrations
codemie codebase ui      # Start and open Codebase Memory graph UI
codemie doctor           # Health check and diagnostics

For a full command reference, see the Commands Documentation.

Codebase Memory MCP

CodeMie can install and orchestrate codebase-memory-mcp with its graph visualization UI:

codemie install codebase-memory
codemie-code init codebase-memory
codemie codebase ui

Use codemie codebase start|stop|status to manage the UI process, or codemie codebase open to open the URL only.

Connect VS Code BYOK via CodeMie Proxy

Configure VS Code's chat language model provider from the active SSO-backed CodeMie profile:

codemie proxy connect vscode

Use --profile <name> for a one-run profile override or --insiders for VS Code Insiders. The command writes that profile's model ID into chatLanguageModels.json and starts or reuses a transparent CodeMie proxy. The profile's project context is passed separately as X-CodeMie-Project.

No plaintext key or generated placeholder is written to that file. On first setup, press ⇧⌘P on macOS or Ctrl+Shift+P on Windows/Linux, run Chat: Manage Language Models, right-click CodeMie Profile Model, choose Update API Key, and enter the local key codemie-proxy. VS Code keeps it in secret storage. Reload VS Code after configuration.

Connect Claude Desktop via CodeMie Proxy

Use Claude Desktop 3P through CodeMie proxy routing to capture claude-desktop metrics and synced conversations.

Prerequisites

• codemie installed
• a valid CodeMie SSO profile
• Claude Desktop 3P installed

1. Connect Claude Desktop

codemie proxy connect desktop

2. Restart Claude Desktop

Quit and reopen Claude Desktop after the proxy configuration is written.

3. Inspect and troubleshoot

codemie proxy status
codemie proxy inspect desktop --limit 5
codemie proxy stop

If Claude Desktop was already using Anthropic subscription or another Gateway

1. Quit Claude Desktop.
2. Sign out or disconnect the previous Anthropic or Gateway provider setup in Claude Desktop.
3. Run codemie proxy connect desktop.
4. Reopen Claude Desktop.

CodeMie cannot safely log you out from Claude Desktop automatically. If the old provider still appears active, clear it in Claude Desktop first and then reconnect through CodeMie.

Documentation

Comprehensive guides are available in the docs/ directory:

• Configuration - Setup wizard, environment variables, multi-provider profiles, manual configuration
  • CODEMIE_INSECURE=1 — disable SSL verification for self-signed certs or local dev environments (SSL is on by default)
• Commands - Complete command reference including analytics and workflow commands
• Analytics Report - HTML dashboard: all 8 views, filters, session drill-down, cost and efficiency metrics
• Agents - Per-agent detail (built-in, Claude Code, Claude Code ACP, Gemini, OpenCode; Pi and Kimi are documented in this README)
• Authentication - SSO setup, token management, enterprise authentication
• Plugins - Reusable packages of skills, commands, agents, hooks, and MCP servers (.claude-plugin/plugin.json format)
• Skills - Injecting custom knowledge into agents via markdown + YAML frontmatter
• Hooks - Shell and LLM-based hooks at agent lifecycle points
• Examples - Common workflows, multi-provider examples, CI/CD integration
• Configuration Architecture - How configuration flows through the system from CLI to proxy plugins
• Proxy Architecture - Proxy plugin system, MCP authorization flow
• Claude Code Plugin - Built-in commands, hooks system, and plugin architecture

Contributing

Contributions are welcome! Please read our Contributing Guidelines to get started.

License

This project is licensed under the Apache-2.0 License.

Links

• GitHub Repository
• Issue Tracker
• NPM Package