Command reference

Find the command for your task, then check its flags. Replace angle-bracket values with your own values. All commands accept --help. The reference tracks the pinned CLI registry, including the three hidden compatibility and administration commands.

init

e11 init

Prepare a fresh or cloned checkout. Start here for project identity, the Toolkit pin, price list, profiles, ignore rules, editor skills, and the knowledge base. Run again to restore generated files; on a checkout that already answered, init then shows the e11 upgrade table.

--knowledge answers what agents should know about and saves it in .foundry/project.toml: none, own (this codebase), e11 (this codebase and the Engineering11 platform), other (picked on a terminal), or other=<name,...>, where a git URL or @-scoped package is added as typed and any other value is a known repository name or path. Get started walks each answer.

Flag	Use
-y, --yes	Take the defaults without asking
--ide <codes>	Comma-separated IDE codes to write skills for (remembered on this machine)
--output-folder <path>	Where planning and stories live, relative to the checkout (default: foundry)
--project-name <name>	The project name in the configuration (default: the directory name)
--knowledge <answer>	What agents should know about: none, own, e11, other or other=<name,...> (saved in the project file)
--from <tarball>	Pin the Toolkit from a local package tarball instead of the registry
-d, --dir <path>	Project directory (default: current working directory)
--no-register	Do not add this project to the dashboard registry
--json	Emit the report as JSON instead of text

doctor

e11 doctor

Check run prerequisites without writing files. Follow fix rows and review rate differences before starting work. A checkout whose knowledge answer is none hears one knowledge-base line and no Alloy rows.

Flag	Use
-d, --dir <path>	Project directory (default: current working directory)
--json	Emit the checks as JSON instead of text
--fix	Apply each row's fix (the init converge, the pinned Alloy commands), then check again

upgrade

e11 upgrade

Show the Foundry CLI, its skills, Alloy and the E11 knowledge pin against latest, and bring them up to date. The knowledge pin changes committed files, so it moves only when named or separately confirmed. Upgrading explains each row.

Flag	Use
--check	Show the table and write nothing; exit 0 current, 1 behind, 2 latest unknown
--json	Emit the report as JSON; without --yes this is --check --json
-y, --yes	Apply foundry, skills and alloy without asking; the knowledge pin moves only with --only knowledge
--only <components>	Restrict to foundry, skills, alloy, knowledge (comma-separated)
-d, --dir <path>	Project directory (default: current working directory)

status

e11 status

Read installed modules and versions. Use doctor to check readiness for a run.

This command has no additional flags.

initiative

e11 initiative <action> [subject]

Use new to create an effort, use to select the current effort, open for a separate worktree, list for available efforts, and show for identity and location. Board actions create, import, or show a managed board. Boards lists the initiative’s boards. Import with --legacy is an upgrade path.

Flag	Use
--slug <slug>	new, use, open, show: the initiative, when not given as the subject
--title <text>	new, board create, board import: the title
--initiative <id>	board actions: the initiative (default: the one this checkout points at)
--location <path>	board create, board import: where the board lives, relative to the initiative directory
--stories <ids>	board create: comma-separated story ids, each optionally id=key
--legacy <path>	board import: the legacy status file, relative to the checkout root
--artifact <id>	board show: the board artifact
--operation-key <key>	board create, board import: the idempotency key (default: a fresh one)
-d, --dir <path>	Project directory (default: current working directory)
--json	Emit the result as JSON instead of text

board

e11 board <action> [story] [status]

Use show to read status, set to change one story, add for new entries, create for a new board, and reconcile after a merge changes the file. Each write uses an operation key.

Flag	Use
--initiative <slug|id>	The initiative (default: the one this checkout points at)
--artifact <id>	The board, when the initiative holds more than one
--title <text>	create: the title
--stories <ids>	create, add: comma-separated story ids, each optionally id=key
--location <path>	create: where the board lives, relative to the initiative directory (default: board.yaml)
--operation-key <key>	set, add, create, reconcile: the idempotency key (default: a fresh one)
-d, --dir <path>	Project directory (default: current working directory)
--json	Emit the result as JSON instead of text

login

e11 login

Sign in to the E11 backend with a device code: the CLI prints a code and opens the approval page in a browser. Already signed in, it says so and changes nothing.

Flag	Use
--no-browser	Print the approval link instead of opening a browser
--force	Sign in again even if already signed in
--json	Emit the result as JSON instead of text

logout

e11 logout

Sign out of the E11 backend and remove the stored token from this machine.

Flag	Use
--json	Emit the result as JSON instead of text

whoami

e11 whoami

Show the account and organisation this machine is signed in as.

Flag	Use
--json	Emit the result as JSON instead of text

registry

e11 registry <operation>

setup (the only operation): set up access to the @engineering11 packages in this machine's user-level .npmrc, using the signed-in account's registry credential.

Flag	Use
--json	Emit the result as JSON instead of text

run

e11 run <kind> [paths...]

Start from story, stories, series, epic, or manifest. Each source compiles to a sealed plan. Review with --plan-only before starting a worker. An epic is a source for the same engine.

Run sources and acceptance explains normal execution.

Flag	Use
-e, --epic <id>	With epic: the epic to select from the epics file
--initiative <id>	The initiative the work belongs to, when the documents sit under no initiative file
--profile <id>	The trusted execution profile to plan with (default: default)
--model <provider/model>	The priced runtime to try first (default: the first priced one)
--max-cost-usd <usd>	Run ceiling. Default: planned items times one call allowance, plus 25 percent headroom.
--max-cost-usd-per-attempt <usd>	The most one worker call reserves (default: the binding’s token bound at the first runtime’s list rate)
--max-turns <n>	The turn allowance of one worker invocation (default: 400). a worker at its allowance is resumed to finish
--review-rounds <n>	The rounds of review a delivery gets before it is accepted or held for a person (default: 0, none)
--plan-only	Print the sealed plan and start nothing
--no-publish	Plan with publication off: no board entry moves and no document is published
--no-report	Narrate no report when the run settles. the engine’s own brief is the report
-d, --dir <path>	Project directory (default: current working directory)
--json	Emit the result as JSON instead of text

runs

e11 runs <action> [run]

Always supply an action. List, status, and report inspect work. Resume and retry continue work. Approve, reject, and adopt record decisions or manual delivery. Stop parks a run. Cancel ends a run. Land records accepted work on the target, cleanup releases its worktree, and successor seals a fresh plan with policy changes.

Adopt, land, and cleanup explains completion.

Flag	Use
--reason <text>	With stop, cancel, retry and reject: why, recorded with the request
--work-item <id>	With retry, approve, reject, adopt and land: the work item
--by <principal-id>	With approve, reject and adopt: who is answering for the work
--commit <sha>	With land: the commit the reached the target ref (default: the accepted head). With adopt: the commit carrying the work (default: the run branch head)
--merge	With land: fast-forward the checkout onto the accepted commit first. the branch the run came from must be checked out and clean
--change <json>	With successor: the policy change as JSON, applied over the original policy
--raise-ceiling <usd>	With resume: raise the spend ceiling to this amount first, then drive on. a run paused for money continues
--approved-by <principal-id>	With successor and resume --raise-ceiling: who approved the change (default: this machine’s principal)
-d, --dir <path>	Project directory (default: current working directory)
--json	Emit the result as JSON instead of text

dashboard

e11 dashboard

Open the operator page on port 4711. The server binds to loopback and offers commands to copy. Register additional checkouts with project add before serving all projects.

Flag	Use
-p, --port <n>	Pin the port (default: 4711, the port e11 run names)
--no-open	Print the url without opening a browser
--all-projects	Serve every registered project in a read-only view.
--principal <role>	Act as this role for one invocation (observer, planner, ux, engineer). overrides .foundry/principal.yaml
-d, --dir <path>	Project directory (default: current working directory)
--json	Emit {url, port} as JSON instead of text

install

e11 install

The earlier name of init. It prints e11 install is now e11 init; running init. and runs init with the flags mapped. Use e11 init in new scripts.

Flag	Use
-y, --yes	Take the defaults without asking (init --yes)
--tools <ids>	Comma-separated IDE codes to write skills for (init --ide)
--directory <path>	Project directory (init --dir)
--output-folder <path>	Where planning and stories live, relative to the checkout (init --output-folder)
--no-register	Do not add this project to the dashboard registry (init --no-register)
--knowledge <answer>	What agents should know about (init --knowledge)
--modules <modules>	Retired: refused with what replaced it
--custom-content <paths>	Retired: refused with what replaced it
--action <type>	Retired: refused with what replaced it
--user-name <name>	Retired: refused with what replaced it
--communication-language <lang>	Retired: refused with what replaced it
--document-output-language <lang>	Retired: refused with what replaced it
-d, --debug	Retired: refused with what replaced it

A retired flag exits 3 before anything is written, and each refusal names what replaced it.

project

e11 project <action> [target]

Register a checkout with add, inspect registrations with list, or remove a registration with remove. Removing a registration changes the dashboard list.

Flag	Use
-n, --name <name>	With add: display name (defaults to the directory name)

agent

e11 agent [agentName] [task...]

Invoke an installed agent through an AI runtime. Use --list to inspect installed agent names and --print to review a prompt. Use run for story execution with run acceptance and accounting.

Flag	Use
-r, --runtime <runtime>	AI runtime: claude | opencode | print | auto (default: auto)
-l, --list	List the agents available in this project and exit
-p, --print	Print the assembled prompt without launching a runtime
-d, --dir <path>	Project directory (default: current working directory)

toolkit

e11 toolkit <verb> [args...]

Skills call these helpers for activation, knowledge techniques, configuration, rosters, rendering, and checks. Verbs include activate, brain, customizable-skills, git-evidence, personas, pick-methods, render, roster, spine-lint, and word-metrics.

Flag	Use
--workspace <dir>	With spine-lint: the folder holding the spine document
-o, --output <file>	Write the result to a file instead of stdout
--project-root <dir>	The project root a verb resolves against (default: current directory)
--skill <dir>	The skill directory a verb resolves against
--skills-root <dir>	With customizable-skills: ssupports this root instead of the installed platform roots
--roster <id>	With roster: resolve this group instead of the default room
--group <id>	With roster: an alias for --roster
--list-groups	With roster: list the groups without resolving any of them
--file <path>	With brain: the technique catalog CSV
--extra <path>	With brain: a JSON overlay of extra techniques, merged into every command
--category <name...>	With brain list and brain random: restrict to these categories
--all	With brain list: dump the whole catalog on purpose
-n <count>	With brain random: how many to draw (default 1)
--json	Emit the result as JSON when supported
--initiative <slug>	With activate: resolve against this initiative instead of the one the checkout points at
--require-initiative	With activate: refuse when the checkout points at no initiative
--range <rev-range>	With git-evidence: the commit range to read, such as main..HEAD
--story <ids>	With git-evidence: the story ids to attribute commits to, comma-separated

visual

e11 visual <action> [args...]

The optional visual verification add-on: before / after / design-reference screenshots of a change with numbered callouts, configured per app by visual.config.json. Actions: run, init, detect, doctor, capture, compose, publish, sheet, schema, screens, clean, and admit. run is the one entry point; the others repair and inspect. admit checks, without taking a slot, whether the machine would admit a web visual run now and prints the reading as one line; run makes the same check before it serves anything, waits while another run holds what it needs, and defers with E_DEFERRED and a resume command when memory or CPU pressure persists. web.streamUrls lists URL globs of requests that never close, left out of the network-quiet wait (the Firestore listen channel always is); web.identities.<name>.lease lets one run at a time sign in as that account, keyed by its credentials and the config's backend. The first doctor (or init) installs a pinned browser runtime into Foundry Home; clean --runtime removes it. See Illustrate a UI change.

The config's top-level evidence sets the app's default evidence: annotated (the default) for numbered callouts and a legend on every composite, or frames for plain frames. An explicit request for one change wins. detect and run print the resolved value as outputs.evidence.

Flag	Use
--spec <file>	The execution spec or SPEC.md whose Screen states name the screens
--screens <file>	A per-run screens file; wins over the spec
--slug <slug>	The run slug when no spec carries one
--out <dir>	The run folder; pass the same folder on every call for one slug (default: {implementation_artifacts}/visual/<slug>)
--app <dir>	The app folder holding visual.config.json (default: visual_app in central config, else the one app found)
--stage <name>	With run: before launches the base and reference captures and returns
--unattended	Never ask, probe, initialise or install; a missing app config is skip no-config
--resume	Re-shoot only screens whose capture is missing or stale
--save	With run: merge the screens into the app config's saved screens file
--json	Print the result envelope as JSON
--ref <ref>	With capture: head, base or reference
--base <ref>	The branch the change merges into (default: origin HEAD, then main or master)
--refresh	With init: probe again and report how the current config differs
--pr <n>	With publish: the pull request number
--reviewed <file>	With publish: the composites/review.json the cold review wrote
--folder <name>	With publish: the folder on the images branch (default: pr-<n>)
--note <text>	With publish: where the images' data came from and what is masked
--no-heading	With publish: leave the heading out of the printed markdown
--dry-run	With publish: render the composites (capturing the base when Before needs it) and print the markdown without pushing
--from <folder>	With sheet: composites, after, before or reference; with run --rebaseline: the git ref to capture the baseline from
--files <list>	With detect and run: touched files, comma-separated, for the app's screenMap
--regenerate	With compose: rebuild every spec, dropping legend edits
--keep-start	With compose: keep the current after frames as the run's start state; a kept screen is never replaced
--rebaseline	With run: replace the branch baseline with this run's after frames, or with --from <ref> a capture of that ref's merge-base with HEAD
--login-vars <vars>	With init: EMAIL_VAR,PASSWORD_VAR to sign in with
--runtime	With clean: remove the browser runtime too
--no-wait	Fail E_LOCKED at once, naming the holder, instead of queueing behind another run for a lock

Every action prints one envelope: { "status": "ok" | "skip" | "fail", "code"?, "remedy"?, "detail"?, "outputs" }. The exit code is 0 for ok and skip, 2 for a usage error, and 1 for any other failure. A skip carries outputs.reason: env-off, config-off, spec-off, disabled, no-config or no-screens. A failure's code is one of E_CONFIG, E_NO_CONFIG, E_SCREENS, E_DEPS, E_LOCKED, E_DEFERRED, E_PRESSURE, E_CEILING, E_SETTLE, E_PORT_BUSY, E_INSTALL, E_BUILD, E_SERVE_TIMEOUT, E_LOGIN, E_BLANK_FRAME, E_ERROR_PAGE, E_MASK, E_FONTS, E_SCREEN_STALE, E_REFERENCE_MISSING, E_INCOMPLETE, E_GEOMETRY, E_PUBLISH, E_UNREVIEWED, E_FINDINGS, E_USAGE or E_INTERNAL, and its remedy says what to do.

errors

e11 errors <action> [args...]

Error intake and the fails-first gate, driven by the e11-errors skill and configured per project in .foundry/custom/e11-errors.toml. Actions: init, ledger, stage, verify-repro, verify-tests, verify-lint, port, nightly, and schedule. init <project> sets a project up and adds a bugfix profile; with --schedule it also installs the nightly job. ledger validates the working ledger, kept under Foundry Home for the checkout's workspace and never in the checkout, prints where it is, and with a flag pulls from the error source, appends rows, cuts a ledger branch or prints the sweep's work order (--queue). stage <id> builds one root cause's story branch, errors/<project>/story/<slug>, off the base: a single commit holding the story, a one-item manifest and the reproduction. verify-repro is the gate the bugfix profile runs: the reproduction must fail with its recorded signature at the staged base and pass at HEAD. verify-tests --kind jest -- <command> is the profile's test check: it runs the team's test command and, when a test fails, holds each failing test against the same command's outcomes at the staged base, so only a test that did not already fail there blocks the fix; without --kind, or when the output holds no report, the exit code alone decides. verify-lint -- <command> is the same check for lint: only a finding (file, rule and message) the staged base did not have blocks. port <id> rebuilds a landed story as errors/<project>/<slug> on the PR base (pr_base_ref, default the remote's default branch): a reproduction commit then a fix commit, replayed red then green there unless the base is unchanged and the tip is the engine's recorded landing; with --push it publishes a finished port instead: it pushes the branch and opens, or reuses, its draft PR from the text the publish session wrote. ledger --cut --push pushes the cut and opens its draft PR. stage, ledger --cut and port --push warn, and commit anyway, on agent-written text that looks like an email, token or key. nightly runs one scheduled night on this machine under a per-project lock: what earlier nights left unfinished (failed landings, unported fixes), then pull, an unattended sweep, one capped engine run per staged cause within the night's budget, ports, an unattended publish, and the CLI's own push, recorded under Foundry Home; --dry-run runs the preflight, prints the plan and spends nothing. The sweep and the engine runs happen in the night's own worktree beside the checkout (<checkout>.errors-<project>), detached at the fetched base, so the checkout is never moved and may hold work in progress. Every step, agent sessions included, holds the source token and the forge credentials. A cause whose fix branch is already on the remote or has an open PR is skipped. schedule installs, removes (--remove) or prints (--print) the launchd job, which runs the CLI itself, and refuses a CLI run from npx's cache; without launchd it refuses and prints the crontab line. Nothing is ever written to the error source.

Every root cause has a slug, a kebab name of at most 40 characters such as expired-token-polling, minted when its ledger row is created and made unique with -2, -3 on a collision. Branches, manifests, commit subjects and every printed line name the cause by its slug; the fingerprint (fp- and 8 hex) is the ledger's join key. Wherever an action takes a root cause (<id>, --fp), it accepts either.

Flag	Use
--project <id>	The error project as configured (default: the only one)
--base <ref>	The branch fixes merge into (default: base_ref in the errors config)
--source <kind>	With init: rollbar or cloudtasks (default: rollbar)
--token-env <name>	With init: the variable holding the source token (default: ROLLBAR_ACCESS_TOKEN)
--environments <list>	With init: comma-separated source environments to read
--pull	With ledger: read the source and pre-classify what is new
--fixtures <dir>	With ledger --pull: read recorded API responses instead of the network
--append <file>	With ledger: validate rows from a JSON-lines file and append them
--cut	With ledger: commit the ledger folder onto a dated branch off the base, leaving the working tree alone
--queue	With ledger: print the order the sweep takes rows in
--triage	With ledger --queue: print the rows a triage session classifies instead
--fp <id>	With verify-repro: the root cause, by slug or fingerprint (default: the branch's first-commit trailer)
--red-only	With verify-repro: in this checkout, require the reproduction to fail with its signature and record it
--harness <dir>	With verify-repro: judge a fix to this package's test runner: red when the base prints no report, green when HEAD reports no failure
--kind <kind>	With verify-tests: jest or playwright, so a failing test already failing at the staged base does not block (default: the exit code alone). With verify-repro --harness: the report the suite command prints
--project-root <dir>	With verify-repro: the checkout holding the ledger (default: derived from the run worktree)
--schedule	With init: also install the nightly job
--remove	With schedule: remove the installed nightly job
--print	With schedule: print the job and the commands an install runs, running nothing
--launcher <file>	With schedule: the program the job runs (default: the first launcher on the nightly PATH that reports this version, else this CLI)
--push	With port: push the finished fix branch and open or reuse its draft PR. With ledger --cut: push the cut and open its draft PR
--dry-run	With nightly: run preflight and print the work order, the run queue and each budget decision, spending nothing
--cause <slugs>	With nightly: a focused run of these causes only (comma-separated): session, run, port, review and push each, skipping triage, recovery and the earlier queue
--first <slugs>	With nightly: put these causes (comma-separated) first in the cause queue for this night; triage, recovery and the rest of the queue run as usual
--sessions <n>	With nightly: start at most this many cause sessions, from the top of the queue
-d, --dir <path>	Project directory (default: current working directory)
--json	Emit the result as JSON

The exit codes are Foundry's own (Error codes and exit codes): 0 for accept, 1 when the action could not evaluate (a runner crash or timeout, an unreachable source, a failed restore), 2 for a refusal or a missing prerequisite (a failed gate, no project root, config, staging commit or ledger row, a dirty tree, a pending restore, a held lock, a missing tool, a platform without launchd), and 3 for invalid input (a usage error, an unknown runner, a budget key that is not positive). nightly exits 0 once its night ran, whatever each step concluded. A refusal's last line is -> refuse <CODE>: <reason>; --json prints { "status": "ok" | "refused" | "error", "code"?, "detail"?, "remedy"?, "outputs" }.

pipeline

e11 pipeline [file]

Run a sequence of agent tasks from .foundry/pipeline.yaml. Steps pass documents through files. Use --dry-run to inspect the sequence and --init to create a starter file.

Flag	Use
-r, --runtime <runtime>	Override the runtime for every step: claude | opencode | print | auto
-d, --dir <path>	Project directory (default: current working directory)
--from <step>	Resume from a step (1-based index or agent name)
--continue-on-error	Keep going past a failing step (overrides per-step continueOnError)
--dry-run	Print the resolved plan without running anything
--init	Write a starter pipeline.yaml from the installed agents and exit

export-plugin

e11 export-plugin

Export selected skills as a local plugin bundle for Claude Agent SDK consumers. Supply a dedicated output directory. Review selected modules and skills before exporting.

Flag	Use
--out <dir>	Output directory for the bundle (required)
--modules <ids>	Comma-separated module IDs to source skills from (default: "foundry")
--skills <names>	Comma-separated skill names to include (default: all skills in the modules)
--name <name>	Plugin manifest name (default: "e11-foundry")
--plugin-version <version>	Plugin manifest version (default: the CLI package version)
--description <text>	Plugin manifest description (default: derived)
--no-clean	Do not wipe the output directory before writing

uninstall

e11 uninstall

Remove generated Foundry installation files from a checkout. Preserve your user artifacts and inspect the resulting Git diff.

Flag	Use
-y, --yes	Remove all E11 components without prompting (preserves user artifacts)
--directory <path>	Project directory (default: current directory)

admin

e11 admin <operation>

Hidden administration commands preview destructive operations under Foundry Home. Operations include detach-project, prune-toolkit-cache, delete-operational-state, and delete-artifacts. Apply requires the current preview’s confirmation token. Follow the operations runbook.

Administration and cutover procedures

Flag	Use
--workspace <id>	detach-project, delete-operational-state: the workspace
--machine <id>	prune-toolkit-cache: the machine (default: this one)
--project <id>	delete-artifact: the project the artifact belongs to
--initiative <id>	delete-artifact: the initiative the artifact belongs to
--artifact <id>	delete-artifact: the artifact
--confirm <token>	Apply the preview the printed this confirmation
-d, --dir <path>	Project directory, for the artifact store (default: current working directory)
--json	Emit the preview or the report as JSON

epic

e11 epic <action> [target] [subject]

Hidden compatibility access for the previous engine. Status, events, and runs read history. Plan refuses and points to run. Dashboard forwards to the current dashboard. Writing actions refuse. Compatibility flags below remain in the parser and do not restore retired writing actions.

Legacy engine compatibility

Flag	Use
-e, --epic <number>	Epic number to select from the epics file (plan/run)
-c, --config <path>	Dependency/verification overlay YAML (epic-run.yaml)
-r, --runtime <runtime>	Worker runtime: claude (live workers) | fake (deterministic replay. default) | print (plan-only)
-d, --dir <path>	Project directory (default: current working directory)
--continue	Keep running stories whose ancestors all succeeded after a failure
-s, --story <id>	Story id for retry (grants fresh attempts) or adopt (records work you finished by hand)
--refresh-spec	Legacy parser compatibility. The retired writing action refuses.
--reason <text>	With stop: why, recorded in the run ledger with the request
--worker-timeout-ms <ms>	Legacy parser compatibility. The retired writing action refuses.
--stall-timeout-ms <ms>	Legacy parser compatibility. The retired writing action refuses.
--max-cost-usd <usd>	Legacy parser compatibility. The retired writing action refuses.
--max-cost-usd-per-story <usd>	Legacy parser compatibility. The retired writing action refuses.
-f, --follow	With status/events: keep streaming new events live until the run reaches a terminal state
--files	With status: list each story's commit subjects and every changed file
-p, --port <n>	With dashboard: pin the port (default: 4711)
--no-open	With dashboard: print the url without opening a browser
--all-projects	With dashboard: serve every registered project instead of this checkout (read-only. see e11 project add)
--revision <sha>	Legacy parser compatibility. The retired writing action refuses.
--verify-only	Legacy parser compatibility. The retired writing action refuses.
--principal <role>	Act as this role for one invocation (observer, planner, ux, engineer). overrides .foundry/principal.yaml
--json	Emit machine-readable JSON instead of the text report
--export	Legacy parser compatibility. The retired writing action refuses.
--rebase-onto-files	Legacy parser compatibility. The retired writing action refuses.

migrate

e11 migrate <area> <action> [subject]

Hidden upgrade commands cover identity bootstrap, storage check, apply, adopt, import, cohort install, cohort complete, and toolkit install. Follow the upgrade checklist for one checkout. Cohort commands belong to the coordinated operations runbook.

Upgrade a checkout

Flag	Use
--cutover <id>	The coordinated cutover this command belongs to (required for storage actions)
-d, --dir <path>	Project directory (default: current working directory)
--file <path>	cohort install: the distributed cohort.json
--workspace <id>	storage apply, storage adopt, storage import: the workspace, one at a time
--consumed <dir>	storage cohort complete: the consumed records collected from every machine
--from <tarball>	toolkit install: pin from a local package tarball instead of the registry
--json	Emit machine-readable JSON instead of the text report

Pinned command surface