Station-Sciences/bot-crossing

What it solves

Bot Crossing is a visual management interface for AI coding agents. It solves the problem of tracking multiple concurrent agent threads across different repositories and different tools (harnesses), providing a spatial map of your agent's activity rather than a traditional list of logs.

How it works

The tool scans local session files from various AI coding agent tools (referred to as "harnesses") such as Claude Code, Codex, and Cursor. It uses a small adapter system to read these files without uploading data or creating accounts.

It represents each repository as a "zone" on a 3D map (Luna, Mars, or Terra). Each agent session is visualized as an "astronaut" and a "building." The state of the agent—whether it is running, errored, or waiting for user input—is indicated by the astronaut's animations and badges (e.g., a ? for unread messages). Users can click an astronaut to open the original thread in its respective harness via deep links.

Who it’s for

Developers who use multiple AI coding agents across various projects and want a high-level, spatial overview of their agentic workflows and the status of their active sessions.

Highlights

• Multi-Harness Support: Supports Claude Code, Codex, and Cursor out of the box.
• Local-First: Reads local files and writes only to a single local colony.json file; no data is uploaded.
• Spatial Memory: Repository zones are "sticky," meaning they stay in the same location on the map to help the user build a mental map of their projects.
• Visual Status Cues: Uses PBR materials, HDR lighting, and specific animations to signal agent states (e.g., hammering for running, sleeping for inactive).
• Deep Integration: Uses OS-level deep links to jump directly from the 3D map into the specific agent thread in the desktop app or CLI.