Hi Product Hunt 👋 I'm Rado, the developer behind Akmon. I built this because I kept running into a specific problem: AI coding tools that work perfectly on a developer's laptop but break completely in the environments where real software actually gets built — CI pipelines, SSH servers, containerized infrastructure, air-gapped networks. The answer was a single statically linked Rust binary with no runtime dependencies. 3.4 MB. Copy it anywhere and it works. The three things I spent the most time on: 1. Provider independence. You should be able to switch from Anthropic to Ollama to OpenRouter without changing how you work. One tool, any model, your keys. 2. The audit trail. Every tool call logged with timestamps, inputs, outputs, and costs. In professional environments you need to know exactly what the agent did. JSONL means you can grep it, pipe it into your observability stack, or just read it. 3. The permission system. Nothing runs silently. Every file write and shell command requires explicit approval before execution. This matters when you are working on production codebases. Still early days. The codebase is solid (~370 tests, clean Rust architecture) but real-world usage always surfaces what tests miss. If you try it I genuinely want to know what breaks, what is confusing, and what you wish it did differently. That feedback is worth more to me right now than upvotes. Happy to answer any questions about the architecture or design decisions.

Akmon

AI coding agent. Single binary. Any model. Full audit trail.

AI coding agent. Single binary. Any model. Full audit trail.