Evidence record / claude-code
A dated record of one change, kept so the writing that cites it can be checked. Compiled from the sources listed below by the research run, not written for reading. The judgment lives in the issues that cite it, below.
2026-09-21-claude-code-effort-model-and-cost-governance-and-claude-plugin-eval
2.1.243 (2026-08-24) adds modelPicker, the modelPricing managed setting, prompt-cache TTL settings and keyless Console sign-in. 2.1.267 (2026-09-09) adds maxEffortLevel, which caps effort on every provider with the lowest cap winning. 2.1.269 (2026-09-11) adds claude plugin eval, which runs a plugin’s eval suite and a no-plugin baseline and reports JSON and HTML. 2.1.274 adds a claude_code.managed_settings_resolved OTel event naming the policy sources a session applied.
Channel: tagged-release (2.1.243, 2.1.267 on stable; 2.1.269 latest only). Half: capability.
Operator consequence: maxEffortLevel in managed settings is the first effort ceiling that binds on Bedrock, Vertex and Foundry. Teams that built a bespoke plugin eval harness can compare against the vendor’s. Wire the managed_settings_resolved event into the fleet dashboard.
Receipt
Finding metadata
Run: 2026-09-21-weekly-digest-2026-08-20_2026-09-21-frontier-v0
Finding ID: 2026-09-21-claude-code-effort-model-and-cost-governance-and-claude-plugin-eval
Profile citations
- Claude Code / claim / effort-model-cost-governance
Source links
Primary links, including exact changelog lines when available.
Versioned source: run artifact