How-to: set the execution mode

Every implementation asks first. At the start of any implementation run (the develop skill, playbook, exec) mokata presents the choice — parallel subagents vs. the sequential gated flow — and never fans out without your pick. The default is the sequential gated flow (lowest cost), and it's asked once per run, not per sub-task.

mokata exec                        # sequential gated flow (default, lowest cost)
mokata exec --parallel             # parallel subagents (fresh-context isolation + review)
mokata exec --parallel --isolation # explicit isolation + two-stage review (E2/E3)
mokata exec --parallel --fanout    # concurrent fan-out (tasks at once)

Run the pipeline in a mode:

mokata playbook              # sequential
mokata playbook --parallel   # parallel (degrades to sequential without a harness)
mokata playbook --parallel --fanout

Save a preference (skip the prompt)

Power users can avoid the per-run prompt with a saved default in the manifest:

mokata config set settings.execution.default sequential   # always sequential, no prompt
mokata config set settings.execution.default parallel      # always parallel (isolation)
mokata config set settings.execution.default ask           # the default — ask each run

ask is the default: always offered, never friction-by-design. A saved sequential / parallel honors the choice without re-prompting.

What parallel guarantees: a token/cost estimate before running (shown when the choice is offered), staying inside the existing gates + audit ledger + token budget, every subagent decision logged, and a clean degrade to sequential — with a clear message — when the harness has no subagents. See execution modes.