project screenshot 1
project screenshot 2
project screenshot 3
project screenshot 4

ConversationStation

Novel agent chat interface lets users interact with their autonomous agents

ConversationStation

Created At

ETHGlobal San Francisco

Project Description

We’ve developed an innovative agent chat interface that enables users to interact with and monitor their autonomous agents in real-time. This project leverages Large Language Models (LLM) and XMTP to create a seamless user interface for querying and commanding agents. Users can ask questions like “What are you built to do?” and the agent will summarize its skills, or ask “What state are you in?” for real-time updates. Users can even prompt actions such as “change parameter X to Y,” and the agent will modify parameters on the fly.

Agents can be customized with distinct personalities and deliver unprompted updates based on specific triggers. A mobile app consolidates all messages and updates from connected agents in one place. By leveraging XMTP, agents can also broadcast updates across multiple social channels, ensuring that critical information reaches users wherever they are.

As autonomous agents become more essential for tasks like trading, selling, gathering information, and developing new actions, our project bridges the gap, enabling smooth interaction between humans and agents. This chat interface will be key in the evolving landscape of human-agent collaboration.

How it's Made

We built this project using the Autonolas framework, which provided the foundation for our agent-based services. The architecture revolves around a set of core skill— Chit_chat — running as individual or multi-agent services that communicate and interact within an agent-economy.

The chat skill integrates LLMs and XMTP: this allows us to bring together a decentralised chat interface for interacting with users, and LLMs which provide context-aware responses.

Technologies Used:

Autonolas framework: The core framework for building, customizing, and deploying the agents.
XMTP: integrating with web3 decetralised messaging.
OpenAI: for providing an LLM to provide contextual responses to users
IPFS: For storing all the code of the agents and their components.
Python: Primarily used for scripting the agents and managing their interactions.
Node: running a websocket interface to XMTP clients
Coinbase: using cdk to connect to a Smart Wallet
background image mobile

Join the mailing list

Get the latest news and updates