Anthropic releases Claude Code v2.1.283 with gateway hints, model controls, and stability fixes
Version 2.1.283 adds gateway request grouping, stricter model allow/deny lists, MCP tool output exposure, and multiple stability and UI fixes across sessions, plugins, and telemetry.
1 source · cross-referenced
- Adds x-claude-code-prompt-id header for LLM gateway request grouping when opt-in via CLAUDE_CODE_GATEWAY_HINT_HEADERS=1
- Introduces availableModelsMatch 'exact' mode and deniedModels setting to restrict model versions
- Exposes MCP tool outputs (WebFetch, WebSearch) and adds OpenTelemetry span events when OTEL_LOG_TOOL_CONTENT=1
- Fixes SDK session state loss, MCP progress notifications, plugin loading, and terminal rendering issues
- Includes /doctor prompt-audit to audit CLAUDE.md files, skills, agents, and commands for older-model prompting patterns
Anthropic published v2.1.283 of Claude Code as an immutable release on GitHub, with changes limited to the release title and notes.
Gateway and control features were added to help operators manage multi-model deployments and correlate requests. The x-claude-code-prompt-id header was introduced so LLM gateways can group requests that serve a single user prompt; enabling requires setting CLAUDE_CODE_GATEWAY_HINT_HEADERS=1.
Two new managed settings were added: availableModelsMatch with an 'exact' mode restricts an availableModels entry to the specific model version it names, preventing new releases from being used until explicitly listed; deniedModels blocks specific models even when availableModels would otherwise allow them.
Observability and debugging were enhanced. MCP tool outputs from WebFetch and WebSearch are now exposed, and OpenTelemetry span events are emitted when OTEL_LOG_TOOL_CONTENT=1 is set.
A new /doctor prompt-audit command audits CLAUDE.md files, skills, agents, and commands for prompting patterns written for older models, helping teams update their configurations proactively.
Stability and correctness fixes span session state, MCP lifecycle, plugin management, terminal rendering, and authentication. SDK sessions no longer lose deferred tool calls or finished tool results when a turn ends early, and MCP progress notifications are preserved when long-running tool calls move to the background.
Plugin handling was improved: stdio MCP servers are cleaned up on session end, stateless remote MCP servers no longer become unusable after a brief HTTP 404, and MCP sign-in no longer fails with an opaque error for servers with no valid URL.
UI and accessibility issues were addressed, including markdown links rendering in Warp terminal, permission dialogs in screen-reader mode, and corrected keybinding documentation and behavior.
Certificate verification for worktree checkouts was fixed when the CA certificate is passed to git via GIT_CONFIG_COUNT environment pairs.
- Oct 14, 2026 · TechCrunch — AI
Science Corporation implants vision-restoring PRIMA chip in first Stargardt disease patient
Trust78 - Oct 13, 2026 · TechCrunch — AI
Flox Intelligence raises $5M and launches edge AI devices to deter wildlife from critical infrastructure
Trust79 - Oct 13, 2026 · Hugging Face
Hugging Face releases LFM2.5-24B-A2B, a 24B-parameter reasoning agent with 2B active parameters for on-device use
Trust79