Claude Code

Confidence 0.90 · 6 sources · last confirmed 2026-06-22

Anthropic’s coding agent — a terminal/CLI-first agent (with a desktop “Claude Cowork” surface) that has become, across the wiki’s 2026 sources, the reference harness practitioners benchmark other tools against. It is mentioned in ~45 wiki sources and is the named instantiation behind much of the agentic-engineering and agent-harness literature: Osmani names it in the convergence observation (top coding agents — Claude Code, Cursor, Codex, Aider, Cline — look more like each other than their underlying models do); Chou cites it as the model of a purpose-built harness (“Anthropic did it wonderfully for Claude Code”).

Its harness primitives recur throughout the wiki: AGENTS.md-style instruction files, hooks, subagents, skills, and — per Vo’s How I AI loops explainer“routines” (scheduled loops) and the /goal primitive (a goal loop: run against a measurable outcome until validated or blocked), shipped at parity with OpenAI Codex’s “automations.”

Appears in this wiki via

Open questions

  • source_count here counts inbound wikilinks; ~45 source pages mention Claude Code in prose without linking. Worth a future lint pass to convert prominent prose mentions to wikilinks now that the page exists.
  • Codex (OpenAI’s coding agent) is the recurring sibling product and is referenced as a [[Codex]] wikilink without a page — a parallel promotion candidate if a second substantive source lands.