Skip to main content
Spine is a multi-agent orchestration platform built on a visual canvas. You describe a project, and a team of AI agents spins up, works in parallel on an infinite canvas, and delivers finished outputs, market reports, financial models, slide decks, interactive prototypes, competitive analyses, SEO audits, and more. The core insight: chat is the wrong interface for complex AI work. Real projects are not linear. They branch, they require multiple streams of research running simultaneously, they need different models for different subtasks. Spine gives both humans and agents a workspace where the structure of the work is explicit and controllable, not hidden inside a context window. Spine is model-agnostic and routes across 300+ models. In a single workflow you can go from an OpenAI LLM call, to an image generation model like Flux or Nano Banana Pro, to Claude generating an interactive app, each block using whatever model fits best.

​
Core Concepts

  • Blocks: Individual AI-powered components that perform specific tasks. Think of them as Lego bricks for AI workflows: each one does something specific, but they can be snapped together and composed in many different ways
  • Canvas: Your infinite workspace where blocks live and connect
  • Context Passing: Information automatically shared between connected blocks via a full dependency graph that determines execution order, parallelism, and concurrency
  • Models: Choose from 300+ AI models, including GPT-5, Claude, Gemini, Grok, Llama, Nano Banana, Flux and hundreds of specialized models. Spine dynamically selects the best model for each task
  • Agent Architecture: A four-tier agent system (L1 / L2 / L2.5 / L3) that decomposes complex projects, coordinates parallel execution, and delivers finished outputs
  • Scheduled Runs: Automate recurring workflows on daily, weekly, or custom schedules
  • Integrations: Connect to 2,000+ apps so agents can read from and write to your existing tools
  • BrowserUse: Agents spin up real browser sessions to interact with the live web
  • Team Workspaces: Add your whole team at no extra cost per person. One shared credit pool, per-member usage controls

​
Modes of Use

Spine supports two primary modes, and most users end up using both. Swarm Mode (Autonomous): Describe a project in natural language via the chat. Agents plan, decompose, execute in parallel on the canvas, and deliver finished outputs. Agents can run autonomously for 80+ minutes on large projects. You can watch them work in real time, or queue a task and come back to finished deliverables. Manual / Hybrid Mode: Work directly on the canvas. Add blocks, write prompts, choose models, make connections, and run blocks yourself. Branch in different directions from the same starting point. You can mix modes freely, let agents do the heavy lifting, then iterate on specific blocks manually.

​
Your First Canvas

  1. Start with a Chat Block: Click anywhere on the canvas and select “Chat Block”
    • Chat with any of the hundreds of models on Spine, change the model mid-conversation if you like
  2. Expand your thinking: Hit the plus icon (or drag and drop an edge from the Chat block) to create new connected blocks. The Chat Block will automatically be used as context for the new blocks
  3. Bring your own docs: Import YouTube videos, files, or previous LLM threads (needs to be publicly shareable) by copying and pasting links or dragging and dropping files. They will automatically show up as blocks on Spine that you can use as context going forward
  4. Let agents take over: Describe a complex project in the chat and let Spine Swarm decompose it, assign persona agents, and deliver finished outputs across multiple block types
  5. See the big picture: Zoom out to view your entire thought process

​
Video Overview

​
Get Involved

Ready to dive deeper? Check out our roadmap to see what features we’re building next and vote on what you’d like to see prioritized. Join our community on Discord to get help, share feedback, and connect with other Spine users.