Orchestration
One lead. A team of agents.
Hand Mel a big task and it runs a whole team. A lead agent decomposes the work, spawns workers across panes, shares memory between them, then integrates and verifies — hands-off.
Workers are Mel agents
The lead spawns worker Mel agents on the models you choose, each in its own pane, coordinating through cross-agent shared memory. Model failures fall back across providers automatically.
Workers are CLI agents
Prefer your installed CLI agents? The lead drives Claude Code, Codex, and opencode in panes — launching them non-interactively and polling each to completion.
The flow
Decompose, spawn, integrate.
- 01Decompose
The lead breaks the goal into independent subtasks, each owning different files.
- 02Spawn
Each subtask gets its own worker agent in its own pane — running in parallel.
- 03Share memory
Workers coordinate through a shared project memory so nothing conflicts.
- 04Integrate
The lead reviews each result, wires it together, and runs the build to verify.