MISSION BRIEFING
IN DEVELOPMENTMAIN DEV2026 · Personal · Public release planned
A local-first terminal client that conducts several AI coding CLIs — Claude Code, Codex, and Grok — as interchangeable workers behind a single conversation. Nexus routes each turn to whichever worker suits the task, chooses the model itself so the decision never lands on the user, compiles permission modes down into each CLI's own launch flags rather than a bespoke sandbox, and briefs a worker on the turns it missed before handing over the thread. It deliberately never calls a model API and never edits a file itself — every action runs through a real worker's CLI, which keeps one audit trail and one permission model across all three.
In development, and not public yet — the plan is to open it once it is out of development. Routing, model pinning, cross-worker memory, permission modes, and resume detection work today; usage-limit failover, the planner, and the importer are not built yet.
This mission is still in active development. Screens and the live URL will be added after launch.
Curious about the internals or want a walkthrough? request a demo.
press ESC to go back