Mistle, open-source infrastructure for running sandboxed coding agents. It focuses on secure credential handling via a proxy, explicit configurations, and allowing users to bring their own models, sandboxes, and agents.

Mistle addresses a critical and emerging need for secure, controlled environments to deploy AI-driven coding agents within enterprise contexts. The explicit design choice to keep credentials out of the sandbox and route access through a proxy directly mitigates significant security risks associated with autonomous code generation and execution. Furthermore, its emphasis on explicit configuration and user-provided components counters the 'black-box' nature often found in AI tools, providing necessary transparency and control for organizations. This open-source offering democratizes access to infrastructure previously developed internally by tech giants, enabling broader adoption of agent-based development workflows while prioritizing security, auditability, and operational clarity. It targets a growing market for AI-assisted software engineering, where trust and control are paramount for enterprise integration.