Station-Sciences/bot-crossing

What it solves

Bot Crossing provides a unified, visual dashboard for managing multiple coding-agent threads across different AI tools. Instead of juggling separate windows or CLI sessions for different agents, it gathers all your active coding threads into a single, interactive 3D map where each repository is a plot of land and each agent is a bot working on that plot.

How it works

It runs locally on your machine as a Vite dev server. It uses small "adapters" to read the local session files of various coding harnesses (like Claude Code, Cursor, or Codex) without ever writing to them. The only file it modifies is data/colony.json, which stores your layout preferences. When you interact with a bot, the tool uses system-level deep links to hand control back to the original harness, ensuring you always use the agent's native interface for actual coding tasks.

Who it’s for

This is for developers who use multiple AI coding agents simultaneously and want a centralized, spatial way to monitor their status, identify which threads need attention, and quickly jump into active sessions without losing track of their work.

Highlights

• Unified View: Displays threads from multiple agent harnesses in one colony.
• Privacy-First: Operates entirely on your local machine; no data is uploaded or stored in the cloud.
• Non-Destructive: It never writes to your agent's session files, only to its own local map configuration.
• Smart Navigation: Bots automatically route around buildings and obstacles using A* pathfinding.
• Status Badges: Bots use visual cues (like a ? for unread threads or ! for errors) to show exactly which sessions require your input.
• Extensible: Adding support for a new agent harness requires only a single new file in the server directory.