Claude Code v2.1.145: JSON Session Export and Plugin Previews
TL;DR
Introduced `claude agents --json` to export live sessions for scripting. Added plugin previews for commands, skills, and servers before installation, and resolved a permission-prompt bypass in Bash commands.
Claude Code has received an update that makes it easier to work with session data. Users can now export their current sessions in a structured format suitable for scripts and automation. This helps keep track of ongoing projects without extra effort.
The new version includes previews for plugins before they are added to the system. People can check out commands, skills, and servers to see what they offer. This way, installations happen with full awareness of the features involved.
A security issue with permission prompts in command line operations has been fixed. The update ensures that prompts appear as expected to maintain control over actions. Overall, these changes support smoother daily use for those building with AI tools.
Who this matters for
- Vibe Builders: Use the new session export to document and share your prompt engineering workflows with your team.
- Developers: Script custom automation pipelines by piping structured session data via the claude agents --json flag.
Harsh’s take
Claude Code is moving from a standalone CLI tool to a composable primitive. The addition of JSON session exports is the most significant change here: it allows developers to treat their AI interactions as data streams for external CI/CD or logging systems. This shift suggests Anthropic wants Claude Code to be the engine inside larger developer platforms, not just a chat interface in the terminal.
The plugin preview feature addresses a growing trust gap in the MCP ecosystem. By forcing visibility into commands and skills before installation, Anthropic is hardening the security model for agentic workflows. This update prioritizes control and auditability, which are the two biggest hurdles for enterprise adoption of autonomous coding agents.
It is a pragmatic step toward making AI agents reliable enough for production environments.
by Harsh Desai
About Claude Code
View the full Claude Code page →All Claude Code updatesMore AI news
- Weekly DigestCursor Composer 2.5 and Claude Code Agent Fixes Lead Weekly Updates
Cursor, Claude Code, and OpenAI Codex each released targeted improvements this period that refine agent behavior, session controls, and integration points for coding workflows across IDEs and terminals.
- App UpdateAdd doctor warnings for hidden MCP server tools
The doctor utility now warns users when sandbox tool policies hide configured Model Context Protocol (MCP) server tools before provider requests are made.
- App UpdateExpose stale-running task maintenance details in CLI
The tasks maintenance CLI command now outputs detailed JSON explanations for retained and reconcile candidates, detailing backing-session, cron, CLI, and wedged-subagent states.