orla

The server's core functionality involves executing arbitrary local commands from the configured 'tools_dir' or 'tools_registry'. While arguments are passed to child processes securely (not directly interpolated into a shell command by Orla), the security of the system fundamentally relies on the user ensuring that the executable tools themselves are trusted, well-audited, and correctly handle their inputs to prevent vulnerabilities like shell injection within the tool's logic. No authentication or authorization is built into the HTTP endpoint beyond the MCP protocol, requiring external security measures like firewalls or proxies for production deployments. The project's `SECURITY.md` explicitly highlights these considerations. The 'avg_tokens_per_call' metric is not applicable as this server executes local commands and does not process LLM tokens itself.