Claude Code 2.0

by Anthropic

Developer-focused agent platform (CLI, VS Code extension, and agent SDK) powered by Sonnet 4.5

See Claude Code docs / Agent SDK notes

Summary

Claude Code 2.0 is a major upgrade to Anthropic’s coding assistant suite. It expands from CLI-first tooling to a full developer UX with a native VS Code extension, improved terminal UX, checkpoints, subagents, background tasks, and a rebranded Claude Agent SDK. It’s aimed at teams that need autonomous agent workflows while keeping safety/rollback guarantees.

Features

  • Native VS Code extension (sidebar, inline diffs, plan mode, auto-accept edits)
  • Improved terminal UX (searchable prompt history, status visibility)
  • Checkpoints: automatic save points of code + conversation context; supports /rewind to revert both code and agent memory
  • Subagents: parallel, modular agent workflows
  • Hooks and background tasks: run tests, lint, long-running servers while agent continues work
  • Claude Agent SDK (formerly Claude Code SDK) for custom agent creation
  • Integration compatibility with VS Code forks and wider dev tooling ecosystems

Superpowers

Designed for developers and engineering teams who want an AI that can work autonomously on multi-step engineering tasks while giving robust rollback and review controls. Checkpoints + plan mode allow risky operations with fast recovery; subagents speed up parallel workstreams.

Pricing

  • Tied to Anthropic model usage (Sonnet 4.5 pricing applies). Enterprise and team pricing likely available—check Anthropic developer portal.

Known limitations & notes

  • Rapidly evolving — some docs may lag behind feature changes.
  • Security and permissions are crucial for agentic features (CI/CD hooks, repo access).
  • Checkpointing complements but does not replace proper VCS and CI workflows.

Sources / notes:

  • Anthropic release notes, community summaries, and documentation for Claude Code and the Agent SDK.