Hi Product Hunt! I’m a UX designer building tools to make working with AI feel more intentional. I built Tern because adding more agents should be a decision, not a reflex. Some tasks are quickest to do directly. Others benefit from a focused second pair of hands. Tern gives your main Codex agent a simple dispatch workflow: decide whether to delegate, choose an available Luna model and reasoning effort, keep context and retries bounded, then check the result. The main agent stays responsible for delivery. It’s free and MIT-licensed. Install from GitHub, start a new Codex task, and say “enable Tern”. You can also activate it for a project through AGENTS.md. I haven’t measured token savings yet, so the website’s token calculator is illustrative. I’d love feedback on which tasks benefit from delegation and which should stay with the main agent.

Tern

The right agent for the job. Less unnecessary work.

The right agent for the job. Less unnecessary work.