Six Environment Variables Silently Void Your Entire Claude Code Policy
2.1.248 added a startup warning for server-managed settings that never arrived. It had to exist because one export in a shell profile skips the fetch — and nothing you configure on the server can undo it.
NeuroX AI · August 28, 2026
The Claude Code 2.1.248 changelog added a startup warning and a /doctor line explaining why server-managed settings weren't fetched. That diagnostic had to be built because six environment variables skip the fetch outright — the five CLAUDE_CODE_USE_* provider flags and any non-default ANTHROPIC_BASE_URL. Until yesterday, nothing told you.
The docs are blunt about the loop you land in: "You can't clear the export with a server-managed env block, because the block arrives through the fetch that the export prevents." An endpoint-managed block doesn't rescue you either — eligibility is checked before managed env blocks apply, so MDM changes which provider the session uses and the fetch stays skipped.
Picture the org that ships Read(./.env) in its deny list plus disableBypassPermissionsMode, verifies it on a laptop, and calls the rollout done. Every engineer running through Bedrock has exactly none of it — and their /permissions output looks perfectly normal.
Three real fixes: unset the export, set the variable to "" in user settings (that applies before the eligibility check), or deliver policy through the endpoint-managed channel or a Claude apps gateway. Note that two of the three depend on the developer you are trying to constrain.
The docs call server-managed settings "a client-side control, not a security boundary." Believe that sentence. A guardrail you have never tried to defeat is a guess.