Claude Code

Editor's Review

Claude Code is Anthropic's terminal-native coding agent, and in 2026 it became a genuine revenue phenomenon — passing a $2.5 billion run rate early in the year. There's no standalone price: it rides on Claude plans (Pro $20/month, Max 5x $100/month, Max 20x $200/month), with usage drawn from a pool shared with Claude chat. The API route is per-token at Sonnet 5's $2/$10 per million input/output tokens — rates Anthropic made permanent on August 11, 2026 when it cancelled a planned price hike. The cautionary tale is Uber: it rolled Claude Code to ~5,000 engineers in December 2025, hit 84% agentic adoption by March, watched per-engineer bills land at $500–2,000/month, and blew its 2026 AI budget by April. Anthropic's own benchmark is ~$13 per developer per active day at API rates — fine for individuals, terrifying at org scale without guardrails. Two stacked caps (a rolling 5-hour session window plus a weekly ceiling shared with chat) mean bursting hard for 2–3 days can empty your week while the session meter still shows headroom. Who should pay? Terminal-native developers who think in diffs and want an agent that reads the whole repo before acting — Claude Code's context-gathering is its superpower. Who shouldn't? Teams without spend controls, or anyone wanting model choice — it's Claude-only by design. Versus Cursor ($20/month, IDE-first, multi-model, best tab completion): Cursor is the better daily driver for most; Claude Code is the better agent for deep, autonomous refactors.