Introducing Phantasm

Thank you for considering using Phantasm. Phantasm is a simple human-in-the-loop (HITL) approval layer for AI agents. With Phantasm, you can build and deploy AI agents that can execute critical business tasks safely.

Most of the information you need to get started can be found in the Repository. This documentation, instead, will focus on explaining the details and concepts in the project. If you have any questions or need help, please feel free to reach out to us on Discord.

Components

Phantasm consists of 3 main components: Client, Server, and Dashboard.

Server

The server is the primary component that connects the client and the dashboard. The server responsible for receiving the approval requests from the client is called the Receiver Server and the server responsible for coordinating the approval requests with the dashboard is called the Coordinator Server.

The Receiver Server uses gRPC while the Coordinator Server uses WebSocket protocol. Both servers are written in Rust and are controlled by Phantasm's CLI. If you're using Docker, this CLI is already the primary entry point for the image.

Client

The client is a language-specific library that you can use to connect your AI agent to the Receiver Server. The core concept of the client is to send an approval request to the server before your AI agent executes a certain function and waits for approval from the approvers.

Dashboard

Phantasm comes with a web-based dashboard that you can use to connect to the Coordinator Server and monitor the incoming approval requests from your AI agents. You or your team can then approve, modify, or reject these requests.