Skip to main content

Agent Setup

Styrby supports 11 CLI coding agents. Install any agent on your machine and Styrby auto-detects it the next time the CLI starts. No extra configuration is needed for detection to work. The agent just needs to be on your PATH.

Pro and Growth both unlock all 11 agents. Agents you do not have installed are simply skipped during detection.

First-Party Agents

Built and maintained by the major AI providers.

Claude Code

Anthropic's official CLI coding agent with deep code understanding and agentic tool-use via the Model Context Protocol. Detected via the claude binary.

# Install
npm install -g @anthropic-ai/claude-code

# Verify
claude --version

Styrby detects Claude Code by checking for claude in your PATH. Configure auto-approve rules under Agents > Claude Code > Permissions in the dashboard.

Codex

OpenAI's CLI coding agent for code generation and understanding, using GPT-4o and reasoning models. Detected via the codex binary.

# Install
npm install -g @openai/codex

# Verify
codex --version

Styrby tracks token usage by intercepting Codex session output. Set your OPENAI_API_KEY environment variable before starting a session.

Gemini CLI

Google's CLI for multimodal AI coding assistance using Gemini 2.0 and 2.5 models. Detected via the gemini binary.

# Install
npm install -g @google/gemini-cli

# Verify
gemini --version

Requires a Google account and Gemini API key. Authenticate by running gemini auth before your first session.

Open-Source Agents

Pro and Growth both run all six of these agents simultaneously with no session limits.

OpenCode

A terminal-based AI coding assistant with multi-provider support, JSON output, and session persistence. Detected via the opencode binary.

# Install
npm install -g opencode-ai

# Verify
opencode --version

OpenCode supports multiple model providers. Styrby tracks whichever model is active in the running session. Check opencode.ai for the latest install instructions.

Aider

AI pair programming in your terminal that works with many LLM providers and integrates directly with Git. Detected via the aider binary.

# Install via pip (recommended)
pip install aider-chat

# Or via pipx (isolated environment)
pipx install aider-chat

# Verify
aider --version

Aider requires Python 3.9 or later. It supports Claude, GPT-4o, Gemini, and many other LLM backends. Set the appropriate API key environment variable for your chosen model provider.

Goose

An open-source AI coding agent by Block (formerly Square) that uses the Model Context Protocol for extensible tool integrations. Detected via the goose binary.

# Install via Homebrew (macOS)
brew install block/tap/goose

# Or via pip
pip install goose-ai

# Verify
goose --version

Goose is Apache 2.0 licensed. Config lives at ~/.config/goose/config.yaml. Check github.com/block/goose for the latest install instructions.

Amp

Sourcegraph's AI coding agent with a "deep mode" that parallelizes code analysis across multiple sub-agents for accurate edits on large codebases. Detected via the amp binary.

# Install via npm
npm install -g @sourcegraph/amp

# Verify
amp --version

Config lives at ~/.config/amp/config.json. Check ampcode.com for the latest install instructions and account setup.

Crush

Charmbracelet's terminal-native AI coding agent with ACP-compatible communication and rich ANSI terminal output. Detected via the crush binary.

# Install via Homebrew (macOS/Linux)
brew install charmbracelet/tap/crush

# Verify
crush --version

Config lives at ~/.config/crush/config.yaml. Check github.com/charmbracelet/crush for the latest install instructions.

Kilo

A community-driven AI coding agent with support for 500+ models and a Memory Bank feature that persists structured project knowledge across sessions. Detected via the kilo binary.

# Install via npm
npm install -g kilo-code

# Verify
kilo --version

Config lives at ~/.config/kilo/config.json. Kilo supports any OpenAI-compatible API endpoint as a backend. Check the agent's official site for the latest install instructions.

Enterprise-Leaning Agents

Pro and Growth both include these. They are listed separately because they target enterprise wedges (Kiro, Droid) rather than the community ecosystem.

Kiro

AWS's AI coding agent with per-prompt credit billing and deep integration with IAM, CodeWhisperer, and Amazon Q Developer. Detected via the kiro binary.

# Install via npm
npm install -g @aws/kiro

# Or download directly from kiro.dev
# Check the official site for platform-specific installers

# Verify
kiro --version

Kiro uses a credit-based billing model. Styrby converts credits to a USD equivalent (1 credit = $0.01) for unified cost tracking. Check kiro.dev for the latest install instructions.

Droid

A Bring Your Own Key (BYOK) AI coding agent that routes to multiple LLM backends through LiteLLM, so you can use Anthropic, OpenAI, Google, Mistral, or on-prem models with your own API keys. Detected via the droid binary.

# Install via npm
npm install -g @droid-ai/cli

# Verify
droid --version

Config lives at ~/.config/droid/config.yaml. Styrby uses LiteLLM pricing tables to estimate costs when the backend does not report token usage directly. Check droid-ai.dev for the latest install instructions.

How Detection Works

When the Styrby CLI starts, it checks your PATH for each of the 11 agent binaries in order:

AgentBinaryConfig File
Claude Codeclaude~/.claude/
Codexcodex~/.openai/
Gemini CLIgemini~/.config/gemini/
OpenCodeopencode~/.config/opencode/
Aideraider~/.aider/
Goosegoose~/.config/goose/config.yaml
Ampamp~/.config/amp/config.json
Crushcrush~/.config/crush/config.yaml
Kilokilo~/.config/kilo/config.json
Kirokiro~/.config/kiro/config.json
Droiddroid~/.config/droid/config.yaml

Detected agents appear in the Agents section of your dashboard. Agents not found in PATH are shown as Not Installed with a link to these docs.

If an agent is installed but not detected, confirm its binary is on your PATH:

# Check PATH for a binary (replace 'claude' with your agent)
which claude

# If not found, confirm the install location and add it to PATH
export PATH="$HOME/.local/bin:$PATH"   # Example for pip installs