How I built a multi-agent simulator as a native SwiftUI app

Description and use

Sorcerer is a free macOS app that simulates AI character interactions without API keys, leveraging existing CLI tools and pure functions for testable logic.

Problem

Develop a free, open‑source macOS app that simulates a community of AI characters reacting to user‑defined hypotheses without requiring API keys.

Outcome

Sorcerer enables rapid, key‑free simulation of AI communities, with testable logic and live streaming feedback.

Workflow

• 1. Prompt user for hypothesis and cast
• 2. Detect user PATH by launching /bin/zsh -ilc "printf '%s' $PATH"
• 3. Select provider CLI (claude, codex, ollama) and construct command with appropriate flags
• 4. Run provider, stream JSON output
• 5. Parse partial JSON to show live typing
• 6. Validate and apply round results via pure functions
• 7. Update world state and display posts, reactions, memories
• 8. Allow forking of world branches

Evidence

I built Sorcerer, a free, open‑source macOS app (SwiftUI, macOS 15+, MIT) that simulates a community of AI characters reacting to a "what if…?" question.

Source

Sorcerer runs the official tools the user is already signed in to: the claude CLI (Claude Code), the codex CLI (ChatGPT), or a local Ollama.

Source

The simulation lives in Engine.swift, whose header says "Prompts and parsing. Pure functions, so the simulation rules are testable without a model."

Source