A cited field guide to Claude Code's 2026 capability stack — the five surfaces, MCP (transports, scopes, OAuth), subagents, the five-type hook system, skills, agent teams — plus the MCP servers worth connecting and a ranked list of projects to build with them.
AI ToolingHermes Agent (NousResearch/hermes-agent) is a real terminal agent CLI with 210K stars, MCP support, and a skills system. It is completely separate from the NousResearch Hermes LLM series.
ToolsPart 1 of 2 — Before comparing Hermes Agent and OpenClaw to Claude Code and OpenCode, you need to know what they actually are. Deep-research findings from 100 agents, primary sources only.
ToolsThese four terminal agents appear in the same lists. Here is what each one is, verified from primary sources, and which to use for what job.
ToolsPart 2 of 2 — The full comparison table and decision guide, built from verified primary sources. Stars, MCP, skills, Windows, free tier, and what each tool is actually for.
ToolsI tested both Hermes Agent and OpenClaw on the same tasks for a month. The answer isn't which one is better — it's that they were built from opposite directions. Here's the architecture that drives everything.
ToolsPart 2: the decision guide, the combined setup, and the honest trade-offs. Based on a month of real testing.
ToolsOpenClaw routes AI through WhatsApp, Telegram, Slack, and 15+ other channels. It also has a skills system and agent loop — and Hermes Agent can migrate from it. Here is what it is.
Tools