macOS alpha, 0.32.0-alpha.1

CLI and shortcuts

Look up command syntax, global options, slash commands and keyboard controls.

A task guide written for this website. Where it summarizes a setting or an API, the Settings reference is authoritative.

Run moguri --help for the exact syntax in your installed build. These references describe the current development CLI. A listed operation still needs its policy, source review, model capability, and runtime readiness.

Choose a launcher

For an installed managed command:

moguri
moguri desktop
moguri --help
moguri status

From a built source checkout:

node bin/moguri.mjs --help
node bin/moguri.mjs status
npm start
npm run desktop

moguri defaults to the TUI. One-shot capability commands connect to the shared service and do not need the desktop UI. Importing the SDK or asking for help does not itself start model inference.

Global options and review values

Option Use
--help Print CLI help before connecting to the service.
--profile PROFILE_ID Use one immutable profile. Overrides MOGURI_PROFILE_ID.
--session SESSION_ID Bind an operation to the actual session.
--project PROJECT_ID Bind a capability to a project. For TUI startup this option takes a project path instead.
--workspace WORKSPACE_ID Select a registered workspace where the operation supports it.
--revision N The revision observed in the current read or preview.
--digest SHA The exact current source or review digest requested by that operation.
--controls FILE Explicit new-session control overrides.
--confirm-review, --confirm-write, --confirm-remove Separate reviewed actions. They are not interchangeable or blanket permission.
--confirm-cost, --confirm-git, --confirm-stage-and-test Specialized cost, source-publication, and validation reviews.
--children archive, delete, or detach The applicable child-tree action, or a saved explicit preference.

Uppercase IDs, revisions, digests, URLs, and filenames are placeholders. Get actual IDs from moguri status or the relevant list/read command. Preview immediately before a write, and do not reuse a digest after the source changes.

Credentials use dedicated UI inputs or piped stdin, never secret shell arguments.

Session and project commands

moguri project add /absolute/path/demo-app
moguri project get PROJECT_ID
moguri project clone REPOSITORY_URL /absolute/new-folder --confirm-write
moguri session create --project PROJECT_ID
moguri session create
moguri session rename 'Demo session' --session SESSION_ID
moguri session mode read-only --session SESSION_ID
moguri session mode agent --session SESSION_ID
moguri session stop --session SESSION_ID
moguri session resume --session SESSION_ID
moguri session retry --session SESSION_ID

Project actions also include sources, diff, porto, repositories, revisioned set and relocate, reviewed repository import/remote changes, and removal with explicit artifact disposition.

Workspace actions include list, discover, create, inspect, reconcile, remove, session, files, and file. They do not adopt or delete arbitrary discovered roots.

Session actions include organize, model-settings, context, spawn, from-pr, from-issue, archive, restore, and confirmed deletion.

Prompt and schedule commands

moguri queue list --session SESSION_ID
moguri queue hold prompt.txt --session SESSION_ID
moguri queue copy QUEUE_ID --session SESSION_ID
moguri queue edit QUEUE_ID prompt.txt --session SESSION_ID
moguri queue run QUEUE_ID --session SESSION_ID
moguri queue remove QUEUE_ID --session SESSION_ID
moguri queue reorder FIRST_ID SECOND_ID --session SESSION_ID
moguri scheduled list
moguri scheduled create timing.json --session SESSION_ID
moguri scheduled send-now SCHEDULE_ID
moguri scheduled cancel SCHEDULE_ID
moguri automations list
moguri automations create workflow.yaml --project PROJECT_ID --paused
moguri automations runs AUTOMATION_ID
moguri automations run AUTOMATION_ID
moguri automations stop RUN_ID
moguri interrupted list
moguri interrupted continue SESSION_ID

Hold does not run the prompt. Scheduled creation consumes an existing saved draft. Automations also take update ID FILE, enable, pause and delete ID [--stop-running]. See drafts and prompt delivery, scheduling, automations and interrupted chats.

Model and budget commands

moguri models list
moguri models refresh
moguri models catalogue
moguri models provider PROVIDER_ID enable
moguri models source API_MODEL_ID CONCRETE_MODEL_ID
moguri models evidence MODEL_ID
moguri models readiness MODEL_ID --session SESSION_ID
moguri session kupo-mode balanced --session SESSION_ID
moguri budget status SESSION_ID
moguri budget set SESSION_ID session-budget.json --confirm-review
moguri dashboard work
moguri dashboard usage
moguri dashboard cost
moguri dashboard velocity

Model commands also include enable/disable, resource-bound token management, runtime preview/review/probe, and readiness-target inspection. Budget commands include dashboard, calls, plans, and separate revisioned profile/global hard policy.

Prices and telemetry are distinct from model admission:

moguri prices
moguri prices refresh
moguri telemetry operations
moguri telemetry storage

Context and retained-document commands

moguri context file /absolute/path/example.txt --session SESSION_ID
moguri context image /absolute/path/example.png --session SESSION_ID
moguri context document /absolute/path/example.pdf --session SESSION_ID
moguri context documents --session SESSION_ID
moguri context document-search ATTACHMENT_ID --session SESSION_ID --filters '{"query":"example","limit":5}'
moguri artifact list
moguri artifact read ARTIFACT_ID
moguri artifact history ARTIFACT_ID
moguri question list
moguri question read QUESTION_ID
moguri plan read --session SESSION_ID
moguri progress --session SESSION_ID
moguri activity list --session SESSION_ID

Context also supports reviewed folder, GitHub, and session captures, original-byte access, bounded document reads, and capture cancellation.

Artifacts support create, update, promote, archive, restore, confirmed delete, exact export, and reviewed project materialization. Question answers require the current question revision. Remote plan saves additionally require the current reviewed binding.

Integration and execution commands

Family Available actions and purpose
github browse, preview, execute, history, read, wait, stop, exact PR ci-preview/ci, and inspection acknowledgement.
linear status, import, catalog, issue, reviewed preview/execute, original history/read/wait/stop, and key management through stdin.
issues, kanban Local tickets, phases, board configuration, paged work items, and explicit continuation tasks.
mcp Session gateway, server metadata, tool discovery, auth status, OAuth login/status/cancel/clear, tokens, and reviewed Remote grants/calls.
remote status, probe, original-run reconcile, inspection acknowledgement, and resource-bound model keys.
sandbox Readiness/runs, bundled setup, reviewed enable/probe/preparation, original execution/result/stop/reconcile, and separate source publication.
worker --stdio Installed versioned SSH worker bridge. It is not a login or provisioning command.
terminal Existing console list/read/reconcile/close, explicit session open, stdin input, owner inspection, and acknowledgement.
firewall Exact Local website allowances. Not an OS firewall for approved shell commands.
tools, tool, tool-result Scoped typed discovery, an exact approved tool job, and original-result pages.
watch, state, coordinate Original owned jobs, canonical service-backed views, and explicit parent/child coordination.
agent-terminal Approved agent command history, not private User-console access.

Native Git commands use the shared slash route:

moguri slash '/git status' --session SESSION_ID
moguri slash '/git diff staged' --session SESSION_ID

Do not infer a standalone moguri git interface from the slash name.

Settings, knowledge, and lifecycle commands

Family Available actions and purpose
profile List, create, revisioned update, explicit retry/stop, and shared-default review.
config, prompt Get/set/reset settings or system prompt. Prompt edits require risk acknowledgement.
memory Index, state, search, and graph on the selected actual root.
skills, plugins Fresh source discovery/read and reviewed imports.
capability Version/history, explicit enable/disable, disabled rollback, and reviewed skill insertion.
knowledge Sky/Terra/Glacier list/search/read/save/update/history/version, explicit promotion/rollback/forget.
hooks, marketplace Reviewed disabled imports and committed catalogue/bundle operations.
cache Exact lock inspection, integrity/advisory preparation, script-free install, registry management, and stdin credentials.
repository Explicit root-bound metadata preview/refresh/state/clear and exact observation stop.
metachat, suggestions, configuration Reviewed settings help, scoped suggestions, and human-approved exact proposals.
attention Read acknowledgement, not question answering or tool approval.
service Explicit restart, stop, or info. info can start a service.
runtime, storage State and owned power status, reviewed store relocation, and explicit power retry.
updates Separate private-release status/history/read/check/cancel and credential handling.
notification-sounds Profile-owned sound metadata/import/removal, not implicit playback.
dictate Status, licensed download, native recording, and local file transcription.
containment Inspect process-group limits or explicitly require qualified Sandbox.

Advanced proposal families are candidate, roles, compound, winddown, decisions, and opt-in agents. They have separate source, cost, stage/test, acceptance, runtime, and publication reviews. A command list is not qualification or permission.

Submit slash commands

Use the actual queued command route:

moguri slash '/model MODEL_ID' --session SESSION_ID
moguri slash '/theme ocean dark' --session SESSION_ID
moguri slash '/read-only' --session SESSION_ID
moguri slash '/kupo status' --session SESSION_ID
moguri slash '/review inspect the current changes' --session SESSION_ID --hold
moguri commands list --session SESSION_ID
moguri commands read EXECUTION_ID

--hold keeps the command inert. Busy requests return their queue identity. UI-only results can wait for a compatible client; they are not silently marked applied.

In a composer, type / to view built-ins and enabled skills. Selection fills or loads the command; it does not automatically run it.

Area Main slash routes
Navigation /session, /sessions, /work-items, /code-requests, /sidebar, /settings, /profiles
Chats /new, /open, /rename, /read-only, /agent, /side, /subsession
Prompts /queue, /run, /steer, /interrupt, /resume, /stop, /copy, /question, /automations, /interrupted
Models /connect, /models, /model, /model-settings, /thinking, /effort, /context-window, /kupo
Review workflows /review, /security, /performance, /debug, /impeccable, /continue, /orchestrate, /spawn, /spar, /critique
Files and documents /files, /file, /attach-file, /attach-image, /attach-folder, /attach-github, /attach-session, /attach-document, /documents, /document-read, /document-search
Access /approve, /approvals, /firewall, /mcp, /mcp-remote, /containment
Work and integrations /git, /github, /linear, /kanban, /work-item, /continue-work, /workspace, /candidate, /compound
Retained information /plan, /progress, /activity, /agent-terminal, /artifacts, /artifact, /dashboard, /budget, /repository
Optional capabilities /memory, /knowledge, /skills, /capability, /hooks, /marketplace, /cache, /compact

Some routes open native views only in a compatible client. /help is the complete catalogue for your build, including aliases and advanced argument forms.

Desktop shortcuts

Shortcut Action
Enter The current composer action, normally Send. Retry or Resume states act on their labeled original operation.
Shift+Enter Newline.
Cmd/Ctrl+Enter Hold during active work; ordinary idle text opens the scheduling/send-now choice. Slash commands remain held.
Shift+Cmd/Ctrl+Enter Explicit send-now or answer from saved composer text.
Cmd/Ctrl+K Command palette.
Shift+F10 on a name Context menu for a project or session.
Alt+Up/Down on a name Reorder within the appropriate sibling group.
Ctrl+Space Default held dictation shortcut after explicit setup.
Left/Right on a focused resize separator Resize the pane; Home/End choose its supported bounds.

Settings editors have their own Save behavior. A save shortcut is not a prompt send. Protected text-editing shortcuts cannot be reassigned to conflicting actions.

Terminal interface

Shortcut Action
Enter / Shift+Enter Composer primary action / newline.
Cmd/Ctrl+Enter Hold active input or use the idle scheduling choice.
Shift+Cmd/Ctrl+Enter Send now; settings editors still use Ctrl+Enter to save.
Ctrl+O / Ctrl+N / Ctrl+P Open project / new session / select session.
Ctrl+L / Ctrl+M Copilot sign-in / model picker.
Ctrl+C Stop active owned work without closing the TUI.
Ctrl+Y / Ctrl+D Approve the exact pending tool / deny.
Ctrl+V Focus full approval arguments.
Ctrl+R / Ctrl+S Reload saved draft / save or keep the local draft explicitly.
Ctrl+G Shared settings and prompt editor.
F2 / /details Expand retained conversation/tool details.
F3 Follow the latest output.
F4 Insert the completed native dictation transcript.
F5 / F6 Artifact library / actual question choices.
F7 / F8 / F9 Dashboard / profiles / Automations, with recurring automations and scheduled prompts.
F10 / F11 Retry the eligible model failure / automatic Kupo mode.
PgUp/PgDn Read history without changing the draft.
Esc / Ctrl+Q Return to composer / save and leave.

Tab focuses navigation. Up/Down chooses a target and Enter opens it. Left/Right operate the sidebar only in navigation focus; composer arrows edit normally.

There is no need to leave a terminal command running just to show a view. Commands and UI actions have retained outcomes, explicit ownership, and their existing policy boundaries.

Back to the user guide