AI Usage Explanation
AI coding assistants were used during development to help draft boilerplate code, structure the WebSocket and MCP command handlers, assist with UI templates, and refine documentation. All generated code and features were thoroughly reviewed, tested, and adjusted manually by the author.
Foundry Browser MCP connects your live Foundry VTT browser session to AI models (such as Claude Desktop, OpenCode, and other MCP-compatible clients) via the Model Context Protocol (MCP).
Through a lightweight local WebSocket bridge (
npx foundry-browser-mcp), this module allows AI assistants to read and manage actors, items, journals, rollable tables, scenes, macros, and compendiums directly in your active world.Key Features:
- Full World Interaction: Create, read, update, and organize entities, roll dice, and switch scenes via AI prompts.
- Custom AI Skills Editor: Create and store private Markdown-formatted instructions directly in your Foundry user account.
- Per-User Security & Controls: Independent WebSocket port settings per user, filtered permissions for non-GM players, and opt-in JavaScript execution (
execute_js). - Built-in Usage Analytics: Track MCP tool frequency, skill usage, and inspect a local audit log of AI actions.
- Multi-Instance Support: Connect multiple Foundry worlds simultaneously using separate bridge ports.
Note: Requires running the companion Node.js MCP bridge (
foundry-browser-mcp on npm) locally alongside your MCP client.