Hi Product Hunt! I'm the solo developer behind leeh. The problem: I work at a large enterprise where all external LLM access is blocked for security reasons. Fair enough — data leakage is real. But we already know how productive MCP-based AI coding tools are. Individuals move ahead, organizations can't keep up. So I asked: what if you use AI tools freely on a separate network, and push only the results in — safely? That's leeh. A one-way push pipeline. Not a dev tool — a delivery tool. It works with whatever AI coding tool you prefer (Claude Code, Gemini CLI, Copilot, Cursor). Key design decisions: - Fail-closed: if a scanner is missing, code is blocked (not skipped) - Server-side policy: the gateway enforces rules, not the client - Human gate: optional admin approval prevents AI agents from pushing unlimited code autonomously Would love feedback on the security architecture. Happy to answer any questions!

leeh

One-way secure push pipeline for AI-generated code

One-way secure push pipeline for AI-generated code