Sorcerer

Sorcerer is a native macOS app that simulates communities of AI agents reacting to what-if questions.

Sorcerer is a native, local-first macOS app that simulates communities of AI agents responding to "what if...?" scenarios. Users summon characters with distinct personalities, relationships, and memories, provide context from local files, websites, or GitHub repositories, and watch discussions unfold round by round. It uses existing subscriptions and CLI tools, supporting the claude CLI, codex CLI, and local Ollama models. The app allows users to pause runs, inject events, conduct polls, pull characters into one-on-one chats, and fork simulation timelines to compare divergent outcomes.

Sorcerer at a glance

Best for: Researcher, Product manager
Runs on: macOS, open source
Pricing: Free and open source under the MIT license; requires existing Claude, ChatGPT, or local Ollama setups.
Released: October 2026
Category: AI agent tool
Website: github.com/gbrasil720/sorcerer

Key features

Simulate agent interactions across feeds, forums, group chats, town squares, and workplaces
Bring your own AI using Claude CLI, Codex CLI, or local Ollama instances
Ingest context from GitHub repositories, local folders, websites, and files
Fork simulation timelines at any round to test alternative scenarios
Conduct off-the-record or public polls across the agent community
Track agent mood, ties, and interactions via world maps and sentiment graphs
Export simulation data as local JSON readable via an in-browser viewer
Customize character personas and avatars using a built-in pixel-art editor

Use it for

Testing community or customer reactions to product pricing changes
Simulating how group chats or social feeds react to breaking news headlines
Exploring narrative and worldbuilding scenarios with interactive fictional characters
Comparing diverging future outcomes by branching agent simulations from a specific round

Alternatives to Sorcerer

BrowserOS — Runs entirely locally and lets AI agents use real user logins to complete tasks.
GOD — Real-time control and intervention for agent simulations.
Merlin — Run Claude Code and Codex on your own machines with your team.
Jahan — Simulates AI populations and traces every turn back to its origin.
Mantid — Trains custom AI models and LLMs directly on your own computer with zero data leaving the device.
Agent Cockpit — It checks evidence chains deterministically so broken claims never slip by unnoticed.
Multipeer — Run AI agents and collaborate with your team live on a single encrypted canvas.
Lancer — It tracks complete task lifecycles with visual branching and real-time activity flows.
AgentGarten — Runs executable code worlds with real-time neural rendering for agents.
Sidekick — Chat with local models using your own files and websites offline.
Pip — It sees your screen and points directly at the exact button to click.
CAMEL — Builds complex multi-agent societies for data generation and simulation.

All Sorcerer alternatives