Tools

Which tools can elevate your operating platform?

AI tooling has three distinct layers. Knowing which layer you're choosing between stops you from comparing incompatible things.

Three Layers

Layer	Definition	Choose when...
Agentic framework	Developer-facing environment + orchestration primitives — where you design and coordinate agents	You want a tool to help you work with AI, compose workflows, and configure how agents behave
Agent engine	Autonomous execution runtime implementing the perception→decide→act loop	You want to deploy an agent that runs independently, wired into your systems
Agent protocol	Coordination standard — between engines, frameworks, and external tools	You need agents and tools to talk to each other reliably

Agentic framework: "where I design and coordinate agents"
Agent engine:      "what runs as the agent in production"
Agent protocol:    "how agents and tools talk to each other"

These are not alternatives to each other. A production setup typically uses all three: Claude Code (framework) to build and test, ElizaOS (engine) to run, MCP (protocol) to connect.

What's Here

Agentic Frameworks

Developer-facing tools with orchestration primitives — sub-agents, task graphs, tool registries, prompt templates. You use these to design agent behavior and run it interactively.

• Claude Code — the most complete setup guide here
• Gemini CLI — Google's CLI with MCP and multi-agent support
• Cursor — IDE-native with agent mode
• Codex — OpenAI's cloud sandbox agent

Agent Engines

Autonomous execution runtimes — they implement the loop and run headless, often wired into external systems. Stateless (per-task) or persistent (long-running daemon).

• ElizaOS — open-source multi-agent runtime
• CrewAI — role-based multi-agent orchestration
• LangChain — composable chain-based agent engine
• Clawbot — deployable always-on personal AI

MCP Servers

Model Context Protocol — the standard for connecting agents to external tools, data sources, and services. Read before choosing any data integration.

AI CLI Tools

Specific-purpose AI-powered command-line utilities. Not orchestration environments — targeted tools for defined jobs.

AI MCP Tools

Curated MCP server implementations — ready-to-wire tools for common integrations.

Context

• Naming Standards — the canonical definitions for agentic framework / agent engine / agent protocol
• Config Architecture — how to configure any of these tools using the agent-agnostic standard
• Agent Protocols — MCP, A2A, Verifiable Intent at the protocol layer
• Agents — when to use subagents and agent teams instead of a full engine
• LLM Comparison — model selection independent of tooling choice

Questions

At what point does adding a dedicated agent engine make more sense than extending your agentic framework — and what's the signal that you've crossed that line?

• If you deploy an agent engine today, what job is it doing that Claude Code running in a loop couldn't do?
• How does the choice of agent protocol (MCP vs A2A vs proprietary) affect which engines and frameworks you can use together?
• What does "stateless runner vs persistent agent" mean for the data your agent accumulates — and who owns that data?