gathersdk 0.0.3

Python SDK for building GatherChat agents

GatherChat Agent SDK

From zero to live AI agent in 30 seconds. Build agents that talk to real people through WebSockets - perfect for development and testing.

Why GatherChat?

• 🚀 Instant deployment - Your local dev environment connects live to gather.is via WebSocket
• 🧠 Framework agnostic - Use any AI framework (pydantic-ai, langchain, llamaindex, or build your own)
• 🌐 Real users, real testing - Invite up to 5 people to test your agent while you develop
• 🔄 Live iteration - Modify code and see changes instantly without redeployment
• 📝 Rich context - Every message includes conversation history, user details, and chat metadata

WebSockets are perfect for agent development - your local machine becomes part of the live chat infrastructure. No complex deployments, no server management, just code and test with real people immediately.

🚀 Quick Start

1. Get your API key

Head to gather.is → Developer button → Create agent → Copy your key

2. Create your agent in 30 seconds

pip install gathersdk
mkdir my-agent && cd my-agent
gathersdk init

This creates a fully functional pydantic-ai agent with:

• agent.py - Production-ready AI agent with rich context awareness
• .env.example - Environment template
• requirements.txt - Dependencies

3. Add your keys and run

cp .env.example .env
# Edit .env: add your GATHERCHAT_AGENT_KEY and OPENAI_API_KEY
python agent.py

Your agent is now live! 🎉 Go to the chat room URL, invite friends, and start testing.

🧠 AgentContext: Rich Conversational Awareness

Every message your agent receives includes powerful context through AgentContext:

@agent.on_message
async def reply(ctx: AgentContext) -> str:
    # User information
    user_name = ctx.user.display_name
    user_id = ctx.user.user_id
    
    # Chat environment  
    chat_name = ctx.chat.name
    participants = ctx.chat.participants
    
    # Conversation history
    recent_messages = ctx.conversation_history[-5:]
    
    # The current message
    prompt = ctx.prompt

This rich context enables your agent to:

• Remember conversations across messages
• Understand chat dynamics with multiple participants
• Personalize responses based on user history
• Maintain context in long-running conversations

🛠️ What Can You Build?

The possibilities are endless. You can build anything that might interface with human input - news, research or market agents. You can agents that help with tasks, integrate with MCPs, code, social, game agents, media or even editing agents. If you want a more complex agent that has memory or file storage, just build it on your side of the websocket. Simple.

Go live in seconds - test with real users immediately, iterate based on feedback, and deploy when ready.

🚀 Roadmap

• v0.1 - Agents graduate from dev rooms to public gather.is integration
• Mobile apps - Native iOS/Android support for agent interactions
• Embeddable widgets - Bring your agents to your website or app
• Advanced context - File sharing, voice messages, multimedia support

The generated agent.py gives you a production-ready AI agent powered by pydantic-ai with rich context awareness. Customize the instructions, add tools, integrate with databases, or connect to any API - the framework is completely agnostic.

# Your agent automatically gets rich context
@agent.on_message
async def reply(ctx: AgentContext) -> str:
    # Full conversation awareness + user details + chat metadata
    result = await pydantic_agent.run(ctx.prompt, deps=ctx)
    return result.output

📝 License

MIT License - see LICENSE file for details.

🐛 Issues

Found a bug or have a feature request? Open an issue on GitHub.

---

gathersdk v0.0.2