Module 3 of 7
3. Dimensions
Three Dimensions of Agency
From simple tool use to autonomous agent networks
Single Agent
One LLM with tools: handles research, coding, data analysis
Examples: Claude + MCP, Cursor, ChatGPT with plugins
Complexity: Simple
Multi-Agent
Teams of specialized agents collaborating on complex tasks
Examples: Researcher + Coder + Reviewer pipeline
Complexity: Moderate
Agent-to-Agent
Agents across organizations communicate via open protocols
Examples: Your booking agent talks to airline agent via A2A
Complexity: Advanced
The Agentic Web isn't just one thing; it can be understood through three interconnected layers or dimensions. Each dimension builds on the one before it, showing how agents evolve from simple reasoners to active participants in a new digital economy.
The Autonomy Spectrum
Four levels of AI agency, from copilot to fully autonomous
L1: Copilot
AI suggests, human decides and executesGitHub Copilot autocomplete80% humanL2: Assistant
AI drafts, human reviews and approvesClaude drafting emails for review50% humanL3: Agent
AI plans and executes, human monitorsAntigravity building a feature end-to-end20% humanL4: Autonomous
AI runs independently, human sets goalsCron-triggered pipeline with error recovery5% human