Skip to content
Tools · Oct 14, 2026

Claude Code v2.1.300 adds virtualText prompt decorations, OpenTelemetry tracing, and AWS-hosted advisor

The update introduces prompt decorations that persist outside the draft, enhanced OpenTelemetry tracing for cache operations, and an AWS-hosted advisor feature on the Claude Platform.

Trust79
HypeLow hype

1 source · cross-referenced

ShareXLinkedInEmail
TL;DR
  • Claude Code v2.1.300 adds virtualText prompt decorations that remain outside the draft and submitted prompt.
  • OpenTelemetry tracing now splits cache writes by lifetime and includes GenAI semantic-convention attributes.
  • An AWS-hosted advisor feature is available on the Claude Platform, controllable via /advisor or advisorModel.
  • Modding API gains $.ui.view({ agentId }) to display an agent’s transcript and prompt.change for prompt box changes.
  • Fixed cloud sessions misreporting organization policy blocks and subagent message loss in /tasks.

Claude Code v2.1.300, released by Anthropic, introduces several developer-focused features and fixes. The update adds virtualText to a mod’s prompt decorations, which displays text in the prompt box that never enters the draft or the submitted prompt. This can be used to show contextual hints or persistent UI elements without altering the conversation history.

OpenTelemetry tracing has been enhanced with two new cache-related fields: cache_creation_5m_tokens and cache_creation_1h_tokens, which split cache writes by lifetime within the api_request event. The tracing suite also adds GenAI semantic-convention attributes to trace spans, including gen_ai.provider.name, gen_ai.conversation.id, gen_ai.tool.name, and error.type. These changes aim to improve observability for debugging and performance tuning.

The modding API now supports $.ui.view({ agentId }), which displays an agent’s transcript in the main view when requested, and $.ui.view({}) to return to the default view. A new prompt.change mod hook is introduced, raised when the prompt box’s text changes without a typed edit—such as from a recalled history entry, a clear action, an undo, or a plugin’s fill. Additionally, the prompt.submit hook can now rewrite a workflow agent’s prompt and model, whereas previously it could only refuse such changes.

The update also adds support for an AWS-hosted advisor on the Claude Platform, accessible via the /advisor command or advisorModel setting. Advisor functionality remains disabled by default and can be toggled off with CLAUDE_CODE_DISABLE_ADVISOR_TOOL=1.

Among the 30-plus fixes, the release addresses cloud sessions incorrectly reporting that an organization policy blocked a host, subagents disappearing from /tasks after typing messages, and the agents view misinterpreting typed navigation keys like "up" or "escape" as commands. OpenTelemetry tool_result events now correctly include git_commit_id after git commit -q in repositories using reftable ref storage. The diff view in Claude Desktop and Remote Control sessions has been corrected to detect conflicted cherry-pick or revert operations in reftable repositories.

Sources
  1. 01GitHub · anthropics/claude-code releases — v2.1.300
Also on Tools

Stories may contain errors. Dispatch is assembled with AI assistance and curated by human editors; despite the trust-score filter, mistakes happen. We correct publicly — every article links to its revision history. Nothing here is financial, legal, or medical advice. Verify before relying on any claim.

© 2026 Dispatch. No ads. No sponsorships. No paid placement. Reader-supported via Ko-fi.

Built by a person who cares about honest AI news.