Skip to content

Changelog ​

All notable changes to this project are documented here. The format follows Keep a Changelog and the project uses Semantic Versioning (pre-1.0: minor versions may contain breaking changes).

Unreleased ​

Fixed ​

  • Internal errors whose code is also a public code (E_BUSY, E_USAGE_LIMIT, …) are now mapped to ClaudeSessionError, so details.reason is always the internal code.

Added ​

  • Idle suspend and transparent resume: idle: { suspendAfterMs }, session.suspend(), session.resume(), status suspended, events suspended and resumed. A suspend never answers or drops a prompt (ICR 0041).
  • First public version of claude-code-driver: drive the interactive Claude Code TUI (pinned to 2.1.316) inside a private tmux server from Node.js, on a Claude subscription.
  • ClaudeSession API: start, send (returns a Turn with events and a done result), command, interrupt, permission answering (respondToPermission, answerQuestion, approvePlan, or a permissions policy/handler), setMode, usage(), state snapshots and typed events, waitFor*, close, closeAll.
  • Observation from hooks (HTTP + unix forwarder in a worker thread), the transcript, the screen, an optional network liveness sidecar, and process exit, reconciled by a pure, replayable core.
  • Fail-closed permissions, sanitized and acknowledged input delivery, owner-death watchdog, orphan reaping, usage-limit wake-up.
  • ccd CLI: lifecheck, view, trace replay|explain.
  • Subpaths claude-code-driver/kernel and claude-code-driver/testing (experimental).
  • Documentation site, architecture guide, contribution guide and security policy.
  • Brand identity (terminal-tile mark, wordmark, favicons, Open Graph and GitHub social cards) and SVG diagrams of the layers, observation channels and step cycle, used across the README, ARCHITECTURE.md and the site (docs/BRAND.md).

Released under the MIT License. Not affiliated with Anthropic.