Anthropic AI Updates: September 12, 2026

1. Claude Code 2.1.269 Adds a Scored Eval Runner for Plugins

Anthropic tagged Claude Code 2.1.269, and the headline addition is claude plugin eval, which runs a plugin’s eval suite against Claude Code and emits scored, reproducible results as JSON plus an HTML report. That turns plugin quality from a vibe check into something a CI job can gate on, which matters now that organizations ship plugins through managed settings. Source

2. The Release Also Opens Up Workflow Concurrency, Output Styles, and Repository-Tagged Telemetry

The same release added CLAUDE_CODE_WORKFLOW_MAX_CONCURRENT_AGENTS (range 1 to 256) to raise the Workflow tool’s per-run agent limit for inference-bound fan-outs, /output-style [name] to list and switch output styles including over Remote Control and in headless and cloud sessions, and OTEL_METRICS_INCLUDE_REPOSITORY to tag OpenTelemetry metrics and events with vcs.* repository attributes. Two smaller additions round it out: the Bash tool result now includes a diff of the files a command changed when it handles file edits (bashEditDiffEnabled), and CLAUDE_CODE_GATEWAY_MODEL_DISCOVERY_TIMEOUT_MS extends the LLM gateway /v1/models discovery timeout past its 3-second default. Source