Spynel - manage ALL of your AI work with one chat.

Spinal: An AI-First Productivity Framework for Multi-Agent Development

This video presents Spinal, an open-source, lightweight framework designed to streamline developer workflows with AI coding agents. Created by the developer for personal productivity, Spinal acts as a centralized orchestration layer that sits on top of existing coding harnesses (e.g., Codex, Agent Zero) without containing any AI itself. It enables developers to manage multiple AI agents across projects from a single chat interface.

Core Architecture & Design Philosophy 🏗️

Spinal operates on three foundational principles:

• One Human, One Chat: The developer communicates exclusively with a single main agent, which handles all delegation and coordination.
• Infinite Leverage: The main agent delegates tasks to an unlimited number of sub-agents, allowing parallel work on hundreds of tasks or complex goals.
• Quality Loops: Every complex task undergoes an automated review cycle within an agentic loop, ensuring quality without manual oversight.

The framework has no embedded AI. It detects the user's existing coding harness, forwards instructions with context, and channels responses back through any connected interface.

Task Management & Agent Pipeline 📋

Spinal implements a folder-based task management system using markdown files with metadata front matter:

• To-Do Folder: The main agent creates detailed markdown task files here when work is required.
• Developer Agent: An internal loop detects new files, spawns a new coding harness session, and instructs the agent to execute the task and update the file.
• Review Folder: Completed tasks move here. A review agent evaluates the work against the original brief and either approves it or sends it back for revision.
• Done Folder: Approved tasks are stored here upon completion.
• Heartbeat Agent: This configurable "janitor" runs periodically (every 15-60 minutes) scanning all folders for stale, broken, or waiting tasks. It can escalate issues through Telegram/WhatsApp for higher responsiveness.

User Interface & Connectivity 📡

Spinal supports multiple communication channels:

• Terminal Interface: A feature-rich text editor with themes, copy/paste support, and full keyboard navigation.
• Telegram & WhatsApp: Full remote access with all slash commands available, including typing indicators and job monitoring.
• Voice Messages: Telegram voice messages are transcribed using an open-source speech-to-text model.
• Slash Commands: Commands like /jobs, /tasks, and /config provide real-time visibility into agent activity and system configuration.

Installation & Initialization ⚡

Setup requires minimal effort:

npm install -g spinal

• Single Instance: Initialize Spinal in a workspace directory to manage all projects through one chat interface, maximizing leverage.
• Per-Project Instances: Initialize separate instances for individual projects, each with distinct configuration and Telegram bot connections.
• Auto-Start: Enable registration for system boot, ideal for 24/7 server management via remote messaging.

Key Takeaway

Spinal transforms the developer-AI interaction paradigm by replacing manual orchestration with automated delegation and quality assurance. Its key innovation is separating the communication layer (you ↔ main agent) from the execution layer (main agent ↔ sub-agents), enabling massive parallelization without overwhelming the developer. The framework's plugin-free design, reliance on existing harness configurations, and remote accessibility through Telegram/WhatsApp make it particularly powerful for developers managing multiple servers or projects while on the go. For those seeking to scale their AI agent usage without added complexity, Spinal provides a practical, zero-configuration solution that prioritizes developer control and workflow efficiency.