Privacy and security
Comprehensive overview of Mind's privacy practices, security measures, compliance standards, and user rights
Mind is built around strict enterprise security infrastructure and follows best practices to protect your development environment and data. Security is a shared responsibility between Mind and you.
URL fetching
In the Mind chat module, you can paste a specific URL for your device to fetch and use it as context to help Mind answer your query or solve your task. You are responsible for the URL content that you fetch and ensuring that your use complies with applicable third-party terms and laws.
Autopilot versus supervised mode
Mind offers two interaction modes, Autopilot and Supervised, that control how you review agent actions. Both modes grant the agent the same capabilities: creating, modifying, searching, and deleting files in your codebase and running commands that impact the filesystem. The difference is the review workflow, not the underlying permissions or access scope.
| Capability | Autopilot | Supervised |
|---|---|---|
| Agent can read files | Yes | Yes |
| Agent can write files to disk | Yes | Yes |
| Agent can run commands | Yes (with policy) | Yes (with policy) |
| Review step before changes applied | No | Yes (diff review) |
Trusted commands
By default, Mind requires approval before running any command. You can create your own trusted commands list:
- Exact matching: Commands must match exactly (e.g.
npm install) - Wildcard matching: Use
*to trust command variations (e.g.npm *) - Universal trust: Use
*alone to trust all commands (use with extreme caution)
Protected paths
Mind requires explicit approval before writing to certain protected paths:
| Pattern | Match type | Description |
|---|---|---|
| .vscode/ | Path contains | Workspace settings |
| .git/ | Path contains | Files inside .git directory |
| .mindignore | Exact basename | Mind ignore rules |
| mcp.json | Exact basename | MCP server configuration |