Skip to content

USAGE.md

Latest commit

 

History

History
156 lines (122 loc) · 5.12 KB

USAGE.md

File metadata and controls

156 lines (122 loc) · 5.12 KB
Raw

Usage

Initialize a new project

Better Agents supports two modes of operation:

Interactive Mode

better-agents init my-awesome-agent

Non-Interactive Mode

Provide all configuration via CLI options and environment variables.

LANGWATCH_API_KEY="sk-lw-..." ANTHROPIC_API_KEY="sk-ant-..." better-agents init my-agent \
  --language python --framework agno --llm-provider anthropic \
  --coding-assistant cursor --goal "Build a trading agent"

The CLI configuration includes:

  1. Programming Language: Python or TypeScript (--language)
  2. Agent Framework: Agno, Mastra, LangGraph, Google ADK, Vercel AI (--framework)
  3. Coding Assistant: Claude Code, Cursor, Antigravity, Kilocode CLI, or None (--coding-assistant)
  4. LLM Provider: OpenAI, Anthropic, Gemini, Bedrock, OpenRouter, or Grok (--llm-provider)
  5. API Keys: Provided via environment variables (see below)
  6. Project Goal: What you want to build (--goal)

Environment Variables

API keys are provided via environment variables, not CLI arguments. This follows security best practices and scales better with multiple providers.

Required for All Projects

LLM Provider API Keys

Set the appropriate variable based on your --llm-provider choice:

Provider Environment Variable(s) Get API Key
openai OPENAI_API_KEY https://platform.openai.com/api-keys
anthropic ANTHROPIC_API_KEY https://console.anthropic.com/settings/keys
gemini GOOGLE_API_KEY or GEMINI_API_KEY https://aistudio.google.com/app/apikey
bedrock AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION (optional) https://console.aws.amazon.com/iam/home#/security_credentials
openrouter OPENROUTER_API_KEY https://openrouter.ai/keys
grok XAI_API_KEY https://console.x.ai/team

Coding Assistant API Keys

Only required if using specific assistants:

Assistant Environment Variable Get API Key
gemini-cli GEMINI_API_KEY https://aistudio.google.com/app/apikey

Examples

Interactive Mode Examples

Python + Agno + OpenAI

LANGWATCH_API_KEY="sk-lw-..." OPENAI_API_KEY="sk-..." better-agents init trading-agent
# CLI will prompt for: language, framework, coding assistant, goal

TypeScript + Mastra + Anthropic

LANGWATCH_API_KEY="sk-lw-..." ANTHROPIC_API_KEY="sk-ant-..." better-agents init customer-support
# CLI will prompt for: language, framework, coding assistant, goal

AWS Bedrock

LANGWATCH_API_KEY="sk-lw-..." \
AWS_ACCESS_KEY_ID="AKIA..." \
AWS_SECRET_ACCESS_KEY="..." \
AWS_REGION="us-east-1" \
better-agents init my-agent
# CLI will prompt for: language, framework, coding assistant, goal

Non-Interactive Mode Examples

Complete Python Project with Agno

LANGWATCH_API_KEY="sk-lw-..." OPENAI_API_KEY="sk-..." better-agents init trading-agent \
  --language python \
  --framework agno \
  --llm-provider openai \
  --coding-assistant cursor \
  --goal "Build an agent that can analyze stock prices and provide trading recommendations"

TypeScript + Mastra + Claude

LANGWATCH_API_KEY="sk-lw-..." ANTHROPIC_API_KEY="sk-ant-..." better-agents init customer-support \
  --language typescript \
  --framework mastra \
  --llm-provider anthropic \
  --coding-assistant claude-code \
  --goal "Build a customer support agent that can handle common queries and escalate complex issues"

LangGraph Python + Gemini

LANGWATCH_API_KEY="sk-lw-..." GOOGLE_API_KEY="..." better-agents init research-agent \
  --language python \
  --framework langgraph-py \
  --llm-provider gemini \
  --coding-assistant cursor \
  --goal "Build a research agent that can gather and summarize information from multiple sources"

AWS Bedrock Non-Interactive

LANGWATCH_API_KEY="sk-lw-..." \
AWS_ACCESS_KEY_ID="AKIA..." \
AWS_SECRET_ACCESS_KEY="..." \
better-agents init my-agent \
  --language typescript \
  --framework vercel-ai \
  --llm-provider bedrock \
  --aws-region us-west-2 \
  --coding-assistant cursor \
  --goal "Build a content generation agent"

Coding Assistant Auto-Launch

After project setup completes, Better Agents automatically launches your chosen coding assistant with a customized initial prompt that includes:

  • Your project goal
  • Framework-specific context
  • Best practices guidance
  • Next steps to get started

The CLI detects which coding assistants are installed on your system and shows installed options first in the selection menu. Not installed assistants appear in gray with "(not installed)" but can still be selected.

You can also select "None - I will prompt it myself" if you prefer to manually launch your coding assistant later with the provided initial prompt.

Development (for contributors)

# Clone the repo
git clone https://github.com/langwatch/better-agents
cd better-agents

# Install dependencies
pnpm install

# Run in development
pnpm dev init test-project

# Build
pnpm build