How do I stop Claude Code users from picking a specific model?

Claude Code 2.1.283 adds a deniedModels managed setting that blocks the models an admin lists. A second managed setting, availableModelsMatch with the "exact" option, makes the allowed-models list match exact model versions, so an organisation can pin approved versions instead of whole model families.

What does /doctor prompt-audit check in Claude Code?

The /doctor prompt-audit command, added in Claude Code 2.1.283, audits CLAUDE.md files, skills, agents and custom commands for patterns written for older models. It is a way to find instructions that may need rewriting after a switch to a newer Claude model.

Can Claude Code send MCP tool output to OpenTelemetry?

Yes, as an opt-in. From Claude Code 2.1.283, MCP tool, WebFetch and WebSearch outputs are added to tool.output OpenTelemetry span events when OTEL_LOG_TOOL_CONTENT=1 is set. Without that variable the content stays out of the spans.

Which bugs did Claude Code 2.1.283 fix?

Claude Code 2.1.283 fixes SDK sessions that lost deferred tool calls or finished tool results, stdio MCP servers left running after a session ended early, stateless remote MCP servers made unusable by an HTTP 404, and a Windows PowerShell tool security issue with drive root deletion. It also shows the weekly Fable limit in /usage again.

Anthropic · 2026-09-25 · major

Claude Code 2.1.283 — admins can block models and audit old prompts

Claude Code 2.1.283 adds a deniedModels managed setting and exact model pinning for admins, a /doctor prompt-audit command for CLAUDE.md and skills, and Amazon Bedrock's Mantle endpoint as a provider.

Admins get tighter control over which models run, and users can check their prompt files for habits written for older models.

Quick facts

Version	v2.1.283
Maker	Anthropic
Released	2026-09-25
New admin settings	deniedModels; availableModelsMatch: "exact"
New command	/doctor prompt-audit
New provider	Amazon Bedrock Mantle endpoint

What is it?

Claude Code 2.1.283 gives organisations two new managed settings: deniedModels blocks specific models, and availableModelsMatch with the "exact" option pins exact model versions. The same release adds /doctor prompt-audit, which scans CLAUDE.md files, skills, agents and commands for patterns aimed at older models.

How does it work?

The model controls sit in managed settings, the admin-set layer that users cannot override. For observability, MCP tool, WebFetch and WebSearch outputs can now go into tool.output OpenTelemetry span events when OTEL_LOG_TOOL_CONTENT=1 is set, and an opt-in x-claude-code-prompt-id gateway header groups LLM gateway requests. A mantle upstream provider targets Amazon Bedrock's Mantle endpoint.

Why does it matter?

Teams that must approve each model version can now enforce that inside the tool instead of by policy memo. The prompt audit arrives as many users move to Claude Opus 5.5, when instructions tuned for older models may work against the new one. The fix list is long too: MCP servers left running after a session, lost tool results in SDK sessions and a Windows PowerShell drive-root deletion issue.

Who is it for?

Claude Code admins and heavy MCP users

Frequently asked questions

How do I stop Claude Code users from picking a specific model?: Claude Code 2.1.283 adds a deniedModels managed setting that blocks the models an admin lists. A second managed setting, availableModelsMatch with the "exact" option, makes the allowed-models list match exact model versions, so an organisation can pin approved versions instead of whole model families.
What does /doctor prompt-audit check in Claude Code?: The /doctor prompt-audit command, added in Claude Code 2.1.283, audits CLAUDE.md files, skills, agents and custom commands for patterns written for older models. It is a way to find instructions that may need rewriting after a switch to a newer Claude model.
Can Claude Code send MCP tool output to OpenTelemetry?: Yes, as an opt-in. From Claude Code 2.1.283, MCP tool, WebFetch and WebSearch outputs are added to tool.output OpenTelemetry span events when OTEL_LOG_TOOL_CONTENT=1 is set. Without that variable the content stays out of the spans.
Which bugs did Claude Code 2.1.283 fix?: Claude Code 2.1.283 fixes SDK sessions that lost deferred tool calls or finished tool results, stdio MCP servers left running after a session ended early, stateless remote MCP servers made unusable by an HTTP 404, and a Windows PowerShell tool security issue with drive root deletion. It also shows the weekly Fable limit in /usage again.

Try it

npm i -g @anthropic-ai/[email protected]