A lightweight, isolated context-routing framework designed specifically for terminal-based AI coding agents (Claude Code, Opencode, Antigravity 2.0, Qwen CLI, and Mimo CLI). It organizes your workspace into isolated environments (e.g., work and individual) to keep agent context clean, minimize token usage, and enforce structured workflows.
Warning
IDE Compatibility: This framework is designed specifically for terminal-based CLI agents that read CLAUDE.md natively on start. It is not fully compatible with IDEs like Cursor or Windsurf out-of-the-box, as they do not support dynamic local rule switching during chat sessions.
This repository contains bilingual documentation (English and Russian). You can safely delete the files in the language you do not need:
- To keep English only: delete all
*.RU.mdfiles. - To keep Russian only: delete the English
.mdfiles and rename the*.RU.mdfiles to take their place (e.g., renameREADME.RU.mdtoREADME.md).
- Clone the repository:
git clone https://github.com/boriskin35/LoomRoute.git
- Navigate into the directory:
cd LoomRoute - Launch your CLI agent (e.g.,
claudeorqwen). - Type
helpto see the list of available commands. - Try creating a new environment: "Create a new environment named learn" or type
helpto see options.
CLAUDE.md: Root constitution (Router agent role).registry.md: Global routing registry.skills/: Global agent skills (archive, safe-copy, etc. in English).templates/: Document templates (briefs, checks).work/: Professional environment (code, web apps, deployments).individual/: Personal environment (Second Brain, diary, ideas).
LoomRoute comes with 9 preconfigured commands. Type them directly into your agent's chat:
help: Lists all available commands.troubleshoot: Initiates systematic step-by-step debugging.review: Audits and reviews your code.note <text>: Captures a quick thought into your Second Brain inbox.notes: Shows recent raw drafts in your inbox.brief <topic>/brief-lite <topic>: Generates Karpati task briefs.checks <topic>: Creates verification checklists.archive: Safely archives completed work packets.
You can easily add your own custom commands (like deploy or test) in three simple steps: write a skill markdown file, register it in skills/commands.md, and run it. For a step-by-step guide, see docs/ARCHITECTURE.md#4-how-to-add-custom-commands.
For in-depth explanations, refer to the following documents inside the ./docs/ folder:
- Ecosystem Architecture: docs/ARCHITECTURE.md (Russian: docs/ARCHITECTURE.RU.md): Philosophy, visual folder tree, and file-reading protocol.
- Method Karpati Workflow: docs/METHOD_KARPATI.md (Russian: docs/METHOD_KARPATI.RU.md): Work packets (
brief.md+checks.md) methodology. - Catalog of Commands: skills/commands.md: Detailed description of available commands.
This project is licensed under the MIT License - see the LICENSE file for details.