CLI flags

Only --no-auto-upgrade is a persistent option inherited by subcommands. Other options belong to the root command or a specific subcommand and must appear after the command that owns them. The removed options -C / --cd, -c / --config, --silent, and run --mode are no longer supported.

Root, run and ACP

Option	Short option	Value	Default	Scope	Description
--verbose	—	—	false	root / run / acp / doctor	Output detailed tool trace or diagnostic information
--logs-path	—	File path	—	root / run / acp / doctor	Diagnostic log file path
--no-auto-upgrade	—	—	false	Persistent; also exposed directly by run	Disable startup update checks
--non-interactive	—	—	false	root / run	Print help under root; disable prompts under run
--resume	—	session ID	—	root / run	Resume the specified session
--continue	—	—	false	root / run	Resume the most recently active session; cannot be used with --resume
--tool-call-permissions	—	read / workspace / auto / yolo	auto	root / run	Tool permission policy
--yolo	—	—	false	root / run	Deprecated; use --tool-call-permissions yolo instead
--json	—	—	false	run	Output JSON events
--output-last-message	-o	file path	—	run	Write run output to file (26.9.11)

run uses Agent mode. In 26.9.11, -o writes renderer output rather than only the final message. Combined with --json, it writes JSONL; see Run mode.

mogick run --non-interactive --json --output-last-message run.jsonl "Check code risks"

Execution mode

In the TUI, press Shift+Tab to switch between Agent and Plan modes. An ACP host switches modes through the protocol. Plan mode supports read-only file and Git exploration, Skill discovery and loading, user questions, task lists, plan reading and writing, and available local code search. It does not allow file writes, shell commands, MCP calls, or other network operations.

After completing a plan, the model calls plan.write and ends the turn. The TUI or ACP host presents the plan for approval and handles the mode switch. Plan mode does not use plan.exit.

mogick --continue

Tool permission policies

Policy	Behavior
read	Automatically allow read-only tools and selected safe tools that change only Mogick's internal state
workspace	Include read behavior and automatically allow non-deletion file.write / file.edit calls inside the workspace
auto	Default policy; allow safe operations automatically and evaluate other calls for approval
yolo	Automatically approve tools that require confirmation; explicit deny still takes effect

Non-interactive execution cannot wait for manual confirmation. In CI, select read, workspace, or—only in a fully isolated environment—yolo, based on task risk.

mogick run --tool-call-permissions yolo "Refactor controller error handling"

Session options

--json is persistent within sessions; sessions feedback ... dislike accepts --reason. Export supports readable history or JSON, not the former context/protocol/Git/subagent export modes. See Session export. Use TUI /rewind for checkpoints; 26.9.11 has no sessions checkpoint command.

Diagnostic options

Option	Description
--json	Output JSON
--format md|json	Explicitly specify the format
--export <path>	Write raw diagnostic report
--check-network	Test network services
--check-terminal-font	Interactively check terminal character display
--all	Perform all extra checks
--verbose	Output detailed diagnostics
--logs-path <file>	Specify log file

Logs and updates

--logs-path receives a file path, not a directory. --no-auto-upgrade disables checking for updates at startup.

mogick --logs-path ./mogick-debug.json --verbose --no-auto-upgrade