Claude Code v2.1.122 adds Bedrock service tier selection and improves MCP server handling
Anthropic's Claude Code releases version 2.1.122 with infrastructure enhancements including AWS Bedrock service tier configuration, pull request session tracking, and multiple bug fixes.
1 source · cross-referenced
- Claude Code v2.1.122 introduces ANTHROPIC_BEDROCK_SERVICE_TIER environment variable for selecting default, flex, or priority Bedrock service tiers via X-Amzn-Bedrock-Service-Tier header.
- PR URL pasting into /resume search now identifies the original session across GitHub, GitHub Enterprise, GitLab, and Bitbucket platforms.
- Release includes 15+ bug fixes addressing OpenTelemetry logging, Vertex AI token counting errors, image resizing for newer models, and voice mode Caps Lock keybinding issues.
- Fixed /branch command that previously generated forks with tool_use ID mismatches when sourcing from rewound timeline sessions.
Anthropic published version 2.1.122 of Claude Code on April 28, with updates targeting infrastructure configuration and bug resolution. The release introduces an environment variable for specifying AWS Bedrock service tiers—default, flex, or priority—transmitted as an HTTP header, allowing users to optimize cost and latency tradeoffs.
The update extends session discovery capabilities by enabling PR URL search across multiple version control platforms. Users can now paste a pull request URL into the /resume search feature to locate the original Claude Code session that created it, functioning across GitHub, GitHub Enterprise, GitLab, and Bitbucket.
The release resolves a critical issue in the /branch command that caused fork generation failures when the source session contained entries from rewound timelines, manifesting as missing tool_use ID blocks. Additional fixes address Vertex AI token counting failures for proxied users, incorrect image resizing for newer models (2576px instead of 2000px), and unresponsive Caps Lock keybinding detection in voice mode.
OpenTelemetry logging received improvements to emit numeric attributes as actual numbers rather than strings, and MCP server connector deduplication now provides hints when duplicate connections are detected.
- Apr 29, 2026 · TechCrunch — AI
Amazon launches OpenAI models and new agent service on AWS Bedrock
Trust56 - Apr 28, 2026 · Hugging Face / NVIDIA
NVIDIA Nemotron 3 Nano Omni adds audio and video capabilities to multimodal AI model
Trust63 - Apr 28, 2026 · GitHub · huggingface/transformers releases
Hugging Face releases Transformers v5.7.0 with new model support and bug fixes
Trust79