Configure Claude Code Auto Mode by setting the permissions field in your settings.json file to "auto" and ensuring your Team plan account has access to the research preview. Locate your settings.json file at ~/.claude/settings.json (Mac/Linux) or %APPDATA%\Claude\settings.json (Windows). Add or modify the permissions section: {"permissions": "auto"}. Auto Mode then becomes available for use. Verify availability: Auto Mode is currently a research preview on Team plans ($30/month per user), with availability expanding to Enterprise and API users. Check your account settings to confirm you have access. Once configured, Auto Mode evaluates each tool call before execution. You don't need to enable it per-session—it operates by default once configured. Optional: Fine-tune Auto Mode behavior by configuring AllowedTools whitelist in settings.json. Even in Auto Mode, you can restrict Claude to specific safe operations. For example: {"allowedTools": ["git", "npm", "python", "node"]} restricts Claude to version control, package management, and language runtimes while blocking dangerous commands like rm, dd, chmod. This is particularly useful in production environments. For team consistency: commit your CLAUDE.md configuration file to your repository (not settings.json, which is personal). Include permission guidelines, MCP integrations, commit conventions, and any required directory structures. All team members benefit from standardized settings. Disable Auto Mode if needed by reverting permissions to "manual" (default) or "dangerous" (skip all checks). Monitor Auto Mode behavior: in VS Code or your IDE, enable verbose logging to see which tool calls the classifier approves or blocks. This helps you understand the safety system and adjust configurations. Auto Mode doesn't require special system permissions or elevated privileges—it operates within your existing user account permissions. When using Claude Code for large-scale development tasks, Zilliz Cloud provides a managed vector database that stores code embeddings, enabling semantic search across your repository—letting the agent quickly locate related code patterns, APIs, and architectural patterns without exact keyword matching.
Learn more: