Claude docs changes for April 16th, 2026 [diff]

Executive Summary

Version 2.1.110 adds the /tui command for flicker-free rendering in the same conversation, mobile push notifications via Remote Control, and fixes MCP tool calls hanging when server connections drop
Agent SDK gains excludeDynamicSections option to move per-session context out of the system prompt, enabling cross-machine prompt cache reuse for fleet deployments
Code Review's REVIEW.md documentation is significantly overhauled — it now acts as the highest-priority system prompt override with detailed guidance on tuning severity, nit volume, skip rules, and repo-specific checks
Two new environment variables (CLAUDE_CODE_REMOTE, CLAUDE_CODE_REMOTE_SESSION_ID) make it easy to detect cloud sessions and link generated artifacts back to their source transcript
claude.ai/code web sessions now support URL query-parameter pre-filling, enabling integrations like issue-tracker buttons that open Claude Code with a pre-loaded prompt and repository

Added /tui command and tui setting — run /tui fullscreen to switch to flicker-free rendering in the same conversation
Added push notification tool — Claude can send mobile push notifications when Remote Control and "Push when Claude decides" config are enabled
Added autoScrollEnabled config to disable conversation auto-scroll in fullscreen mode
Added option to show Claude's last response as commented context in the Ctrl+G external editor (enable via /config)
--resume/--continue now resurrects unexpired scheduled tasks
SDK/headless sessions now read TRACEPARENT/TRACESTATE from the environment for distributed trace linking
Session recap is now enabled for users with telemetry disabled (Bedrock, Vertex, Foundry, DISABLE_TELEMETRY); opt out via /config or CLAUDE_CODE_ENABLE_AWAY_SUMMARY=0

Changed Ctrl+O to toggle between normal and verbose transcript only; focus view is now toggled separately with the new /focus command
Improved /plugin Installed tab — items needing attention and favorites appear at the top, disabled items are hidden behind a fold, and f favorites the selected item
Improved /doctor to warn when an MCP server is defined in multiple config scopes with different endpoints
/context, /exit, and /reload-plugins now work from Remote Control (mobile/web) clients
Write tool now informs the model when you edit the proposed content in the IDE diff before accepting
Bash tool now enforces the documented maximum timeout instead of accepting arbitrarily large values

Fixed MCP tool calls hanging indefinitely when the server connection drops mid-response on SSE/HTTP transports
Fixed non-streaming fallback retries causing multi-minute hangs when the API is unreachable
Fixed PermissionRequest hooks returning updatedInput not being re-checked against permissions.deny rules; setMode:'bypassPermissions' updates now respect disableBypassPermissionsMode
Hardened "Open in editor" actions against command injection from untrusted filenames
Fixed stdio MCP servers that print stray non-JSON lines to stdout being disconnected on the first stray line (regression in 2.1.105)
Fixed PreToolUse hook additionalContext being dropped when the tool call fails

Added an explicit disclaimer that total_cost_usd and costUSD are client-side estimates computed from a bundled price table, not authoritative billing data — pricing changes, unrecognized models, and billing rules can cause them to drift from the actual bill. [line 6] [Source]
Updated diagram description and step 2 caption to say "cumulative estimate" rather than "authoritative total". [lines 29-41] [Source]
Updated troubleshooting guidance to say the result message total_cost_usd is the SDK's best estimate, not a billing-authoritative figure. [line 161] [Source]

Added new section "Improve prompt caching across users and machines" documenting the excludeDynamicSections option (exclude_dynamic_sections in Python). When set, per-session context (working directory, git status, memory paths) moves from the system prompt into the first user message, allowing a fleet of agents on different machines to share the same cached system prompt. Requires SDK v0.2.98+ (TypeScript) or v0.1.58+ (Python). [lines 204-240] [Source]

Added exclude_dynamic_sections optional field to SystemPromptPreset with description of its effect on prompt cache reuse. [lines 829-837] [Source]
Updated costUSD field description to "Estimated cost in USD for this model, computed client-side" with a link to billing caveats. [line 1420] [Source]

Updated systemPrompt parameter type to include excludeDynamicSections?: boolean and added documentation on its use for cross-machine prompt-cache reuse. [line 314] [Source]
Added note to ModelUsage that the costUSD value is a client-side estimate with a link to billing caveats. [line 2258] [Source]

Added new "Link artifacts back to the session" section: each cloud session exposes its ID via CLAUDE_CODE_REMOTE_SESSION_ID, which can be used to embed a traceable link to the transcript in PR bodies, commit messages, or reports. [lines 98-104] [Source]

Clarified that review results are posted as inline comments with a summary in the review body. [line 20] [Source]
Significantly overhauled the REVIEW.md section: it is now described as highest-priority system prompt override injected into every review agent, taking precedence over default guidance. Added detailed "What you can tune" subsection covering severity redefinition, nit volume caps, skip rules, repo-specific checks, verification bar, re-review convergence, and summary shape. Updated example REVIEW.md with a more comprehensive illustration. [lines 137-198] [Source]
Added new troubleshooting section "Review didn't run and the PR shows a spend-cap message" explaining what happens when monthly spend cap is reached and how to resume. [lines 236-238] [Source]

Added disclaimer that the /cost command dollar figure is a local estimate and may differ from the actual bill; links to Claude Console for authoritative billing. [line 13] [Source]

Added CLAUDE_CODE_REMOTE env var — set automatically to true in cloud sessions, readable from hooks or setup scripts to detect the cloud environment. [line 114] [Source]
Added CLAUDE_CODE_REMOTE_SESSION_ID env var — set automatically in cloud sessions to the current session's ID for constructing transcript links. [line 115] [Source]
Added CLAUDE_CODE_TMUX_TRUECOLOR env var — set to 1 to enable 24-bit truecolor inside tmux (disabled by default since tmux does not pass through truecolor escape sequences unless configured). [line 136] [Source]

Clarified that network.allowUnixSockets is macOS only — on Linux and WSL2 the seccomp filter cannot inspect socket paths, so allowAllUnixSockets must be used instead. [lines 271-272] [Source]

Added new "Pre-fill sessions" section documenting URL query parameters (prompt, prompt_url, repositories, environment) that pre-populate a new cloud session — enabling integrations like issue-tracker buttons that open Claude Code with the relevant context already loaded. [lines 134-152] [Source]

No significant API documentation changes today.