deployment
9 questions
- AI
How do I authenticate Claude Code in different environments?
Six methods in priority order: cloud provider env vars, ANTHROPIC_AUTH_TOKEN, ANTHROPIC_API_KEY, apiKeyHelper script, CLAUDE_CODE_OAUTH_TOKEN, and subscription OAuth. Each has different scopes and footguns.
- AI
How do I run Claude Code through my company's Amazon Bedrock account?
Set CLAUDE_CODE_USE_BEDROCK=1 and AWS_REGION, then authenticate with aws sso login or an API key. Model IDs use the us. cross-region prefix. Fast mode, ultraplan, and Remote Control are not available on Bedrock.
- AI
How do I run Claude Code through Azure AI Foundry?
Set CLAUDE_CODE_USE_FOUNDRY=1 with your resource name or base URL. Authenticate with az login for Entra ID or set ANTHROPIC_FOUNDRY_API_KEY for API key auth. Unlike Bedrock and Vertex, Foundry shows errors immediately rather than falling back silently.
- AI
How do I point Claude Code at our Google Cloud Vertex AI endpoint?
Set three environment variables, authenticate with gcloud, and configure gcpAuthRefresh for sessions longer than an hour. MCP tool search is off by default on Vertex. Per-model region overrides use VERTEX_REGION_CLAUDE_* variables.
- AI
What data does Claude Code send to Anthropic, and how do I control it?
Prompt content is always sent for inference. Telemetry, crash reports, and usage counts are separate and can be disabled independently. Enterprise accounts get ZDR. Bedrock, Vertex, and Foundry deployments send no telemetry to Anthropic by default.
- AI
How do I connect Claude Code to our self-hosted GitHub Enterprise Server?
An admin connects your GHES instance once via a guided GitHub App manifest flow. After that, developers start sessions with claude --remote in any GHES checkout. Plugin marketplaces on GHES require full git URLs instead of the owner/repo shorthand that only resolves github.com.
- AI
How do I run Claude Code from a script without it picking up my local hooks and MCP servers?
Add --bare to claude -p to skip all auto-discovered context (hooks, skills, plugins, MCP, CLAUDE.md, auto-memory). Pass only the flags you explicitly need. --bare is the reproducible mode for CI: the same call produces the same behavior on every runner.
- AI
How do I route Claude Code through our LiteLLM proxy for cost tracking and audit?
Set ANTHROPIC_BASE_URL to your LiteLLM endpoint, choose between unified Anthropic format and pass-through provider endpoints, configure authentication with ANTHROPIC_AUTH_TOKEN or apiKeyHelper, and avoid the compromised LiteLLM PyPI versions 1.82.7 and 1.82.8.
- AI
How do I push Claude Code policy to my whole org when we do not have an MDM?
Use Claude for Teams or Enterprise server-managed settings to deliver JSON policy from claude.ai to every developer's session at startup, without deploying files to devices. Configure permissions, hooks, and autoMode from one place.