andeye · agentic coding on your own Mac
Adversarial agentic models plan, build, test and review each other's work in a secure sandbox – so you ship code you can trust.
This very page was planned, built, tested and reviewed by a vibe session.
In development – built on the same open-core values as everything andeye makes.
A vibe session at a glance – separate models, separate incentives, one shared spec.
Vibe andeye
Letting an AI write your code shouldn’t mean accepting whatever it produces. Vibe andeye puts adversarial agentic models between your idea and your codebase: one plans, one builds, and independent critics, testers and reviewers attack the result until it actually holds up.
How it works
The model that writes the code never judges it. Spec critics, testers and reviewers are separate agents with an incentive to find fault.
A planner drafts the spec, a critic attacks it, a builder writes the code – and independent testers and reviewers decide when it holds up.
The sandbox
Agents work in an isolated container with a firewalled network and per-action permission gates – they can build your project, not wander your machine.
Each session runs in a container with only your project mounted – the rest of your Mac simply isn't there.
Outbound traffic is limited to an allowlist – package registries, GitHub, the model API – so nothing phones anywhere else.
Anything that reaches beyond the sandbox asks first – permission is granted per action, not once for everything.
Who it's for
Vibe is for developers first – but we'd also welcome feedback from non-coding experts who know exactly what they want.