Codiluce

•
22
•

Visualize codebase architecture, monitor real-time AI agent activity, and evaluate change blast radius through interactive repository maps and MCP.

Codiluce bridges the comprehension gap in AI-assisted development by translating codebases into zoomable 2D and 3D architectural maps where every connection is backed by verifiable static analysis. Built to operate alongside autonomous coding agents such as Claude Code, Codex, and OpenCode, it renders live agent activity as pixel characters hopping across the files they inspect, edit, or search. By exposing repository flows, dependency hierarchies, Git evolution, and blast-radius tracking over the Model Context Protocol (MCP), Codiluce equips both humans and AI assistants with grounded architectural context, allowing developers to audit agent work in real time, replay past sessions, and safely inspect complex changes locally.

主要功能

01Git history and architectural snapshot playback with diff tracking and commit-level evolution
02Interactive 2D and 3D zoomable maps displaying repository architecture, nested symbols, and file dependencies
03Real-time visual tracking and session replay of AI coding agents hopping across files during execution
04Evidence-backed dependency analysis and hop-by-hop blast-radius calculation for changed or uncommitted code
05Native MCP server integration allowing coding agents to query repository structure, flows, and architectural context
0622 GitHub stars

使用案例

01Evaluating the blast radius and architectural ripple effects of code changes prior to committing or merging
02Grounding AI coding assistants with verified dependency graphs and structural context via MCP to prevent hallucinations
03Monitoring and auditing multi-agent coding sessions in real time to observe which files and flows are being modified