CLI reference

Browse the commands below, or follow the CLI guide to install, sign in, and complete a workflow.

toloka [options] [command]

Help

Use toloka help [command] to print help for the CLI or a top-level command. For any command, append -h or --help, for example toloka dataset fields add --help.

Commands

Command	Summary
`toloka version`	Print the current CLI version.
`toloka config`	Read and write machine CLI settings.
`toloka config get`	Print a CLI setting, or every setting.
`toloka config set`	Write a CLI setting for this machine.
`toloka login`	Sign in through your browser, or store an API key, after verifying access.
`toloka logout`	Remove stored credentials from this machine and end the browser login.
`toloka auth-status`	Show which platform and credential the CLI would use.
`toloka project`	Select and manage the current platform project.
`toloka project list`	List remote projects this key can see.
`toloka project add`	Register a live project and select it.
`toloka project switch`	Select a registered project.
`toloka project current`	Show the project this command would use.
`toloka project url`	Print the current project URL.
`toloka project remove`	Forget a local project alias.
`toloka project untrack`	Remove a pipeline or dataset from local tracking without deleting it.
`toloka project token`	Mint a project-scoped token for agents and scripts; printed once, never stored.
`toloka project tokens`	List live project-scoped tokens.
`toloka project revoke-token`	Revoke a project-scoped token; it stops working immediately.
`toloka project revoke-tokens`	Revoke every live project-scoped token of the project; they stop working immediately.
`toloka project create`	Create a live project and select it.
`toloka audience`	Inspect the expert audience catalog used by node audience filters.
`toloka audience specialization`	Expert specializations (platform qualifications).
`toloka audience specialization list`	List specializations with their id, name, and expert count.
`toloka generation-model`	Inspect the models synthetic generation nodes can use.
`toloka generation-model list`	List generation models with reasoning efforts and maxCostPerAttempt bounds.
`toloka init`	Initialize this project directory and install the agent bootstrap.
`toloka import`	Generate TypeScript source from a live project.
`toloka plan`	Show what apply would change.
`toloka validate`	Run platform spec validation and report matching checks.
`toloka check`	Run the complete local source, type, and test gate.
`toloka apply`	Check source, then make the platform match it.
`toloka secret`	Manage project secrets.
`toloka secret set`	Set a secret value without writing it to disk.
`toloka expert-list`	Manage the project’s expert lists, which nodes with audienceMode service take experts from.
`toloka expert-list list`	List the project’s expert lists.
`toloka expert-list create`	Create an empty expert list.
`toloka expert-list show`	Show a list and how many of its experts are available.
`toloka expert-list rename`	Rename a list.
`toloka expert-list hide`	Hide a list: its experts can’t be added or removed until it is shown again; nodes that use it keep working.
`toloka expert-list unhide`	Show a hidden list again.
`toloka expert-list experts`	List a page of the list’s experts with their status.
`toloka expert-list add`	Add experts to a list, from arguments or a file.
`toloka expert-list remove`	Remove experts from a list.
`toloka expert-ban-list`	Manage the project’s expert ban list: banned experts can’t work on the project.
`toloka expert-ban-list show`	Show how many experts are banned and in which status.
`toloka expert-ban-list experts`	List a page of the banned experts with their status.
`toloka expert-ban-list add`	Ban experts: they can’t be invited and are removed from the work they’re in.
`toloka expert-ban-list remove`	Lift the bans of experts.
`toloka base-model`	Inspect the global base model catalog.
`toloka base-model list`	List base models.
`toloka base-model get`	Show one base model.
`toloka model`	Manage project Models and immutable versions.
`toloka model list`	List up to 100 Models owned by the selected project (first page only).
`toloka model get`	Show one project Model.
`toloka model versions`	List immutable versions of a model.
`toloka model create`	Create project Model metadata without a Model Version.
`toloka model archive`	Archive a project Model.
`toloka model version`	Manage one immutable Model Version.
`toloka model version download`	Download and verify Model Version content.
`toloka training`	Manage project Training drafts.
`toloka training list`	List up to 100 active Trainings owned by the selected project (first page only).
`toloka training get`	Show one Training draft.
`toloka training create`	Create a Training draft.
`toloka training update`	Update a Training draft. Omitted hyperparameters keep their current values.
`toloka training cost`	Validate Training inputs and calculate the maximum charge.
`toloka training run`	Calculate a ready Cost Estimate and create one queued Training Run.
`toloka training runs`	List up to 100 Training Runs (first page only).
`toloka training status`	Show one Training Run.
`toloka training resume`	Resume a Training Run blocked on funds or payment.
`toloka training cancel`	Cancel a Training Run before training completes.
`toloka training metrics`	Read Training Run metrics.
`toloka training logs`	Read sanitized Training Run logs.
`toloka batch-inference`	Answer dataset rows with a base model or Model Version.
`toloka batch-inference run`	Quote a Batch Inference Run, then confirm it to start.
`toloka batch-inference list`	List up to 100 of the selected project’s Batch Inference Runs (first page only).
`toloka batch-inference status`	Show one Batch Inference Run.
`toloka batch-inference confirm`	Accept the quoted maximum charge and start the run.
`toloka batch-inference resume`	Resume a run blocked on funds or payment.
`toloka batch-inference cancel`	Cancel a run; answers already generated are kept and the GPU time used is charged.
`toloka dataset`	Inspect and move dataset catalog and rows.
`toloka dataset list`	List live datasets in the selected project.
`toloka dataset create`	Create an empty dataset in the selected project.
`toloka dataset fields`	Print field declarations inferred from uploaded rows, or add dataset fields from them.
`toloka dataset fields declarations`	Print field declarations inferred from uploaded rows.
`toloka dataset fields add`	Add dataset fields from fields of the uploaded rows, so filters, `--balance-by` and split columns can use their values; uploads after this fill them too.
`toloka dataset load`	Download live dataset rows into a JSONL file, one object of the chosen fields per row.
`toloka dataset upload`	Append rows to a dataset.
`toloka dataset slice`	Freeze the rows of a dataset, with chosen columns read as they were then.
`toloka dataset slice create`	Freeze the rows that match a filter, or every row, so the same slice always selects the same rows.
`toloka dataset slice list`	List up to 100 slices of a dataset, newest first (first page only).
`toloka dataset slice show`	Show the status, filter, frozen columns and row count of a slice.
`toloka dataset split`	Put every row without a part in one, sampled by share; rows that already have a part keep it, so test rows stay test as the dataset grows.
`toloka dataset columns`	List the columns of a dataset, with the parts and roles of split columns.
`toloka pipeline`	Operate on a live pipeline in the selected project.
`toloka pipeline list`	List live pipelines in the selected project.
`toloka pipeline run`	Start a run on an applied pipeline.
`toloka pipeline lint`	Run the pipeline’s lints that have no current result.
`toloka pipeline status`	Report the latest run’s state and item counts.
`toloka pipeline runs`	List recent runs.
`toloka pipeline cost`	Forecast what a run of this pipeline would bill.
`toloka pipeline cancel`	Request cancellation of a run.
`toloka pipeline results`	Report a run’s items.
`toloka status`	Report project health: live counts and pending source changes.
`toloka report`	Preview and send an issue report to Toloka.
`toloka skill`	Discover bundled agent skills and install them into .agents/skills.
`toloka skill list`	List bundled skills as not-installed, current, or stale.
`toloka skill search`	Find bundled skills for a task.
`toloka skill install`	Copy one bundled skill into .agents/skills/<skillId>.
`toloka ui`	Inspect components available for labeling interfaces.
`toloka ui components`	List Task UI components, or print the props, value shape, and example of the named ones.

Global options

boolean

Output the current version.

string

Project root; otherwise search the working directory and parents; init and report use only the working directory.

string

API key override; login stores the key it verifies.Environment fallback: TOLOKA_SCOPED_TOKEN, TOLOKA_API_KEY.

string

Platform base URL override; login saves the selected URL.Environment fallback: TOLOKA_BASE_URL.

string

Target project; overrides toloka.json.Environment fallback: TOLOKA_PROJECT_ID.

Additional help

Credentials resolve in this order, highest first:
--api-key
TOLOKA_SCOPED_TOKEN
TOLOKA_API_KEY
the API key stored by toloka login --api-key-stdin
the browser login stored by toloka login
The platform defaults to https://platform.toloka.ai.

Environment variables

Client settings read by the CLI are listed below. A variable named with secret set --from-env supplies that secret’s value. Installer variables are described in Install.

string

Project-scoped token. Takes precedence over TOLOKA_API_KEY; --api-key takes precedence over both.

string

API key used when no --api-key or TOLOKA_SCOPED_TOKEN is set; takes precedence over saved credentials.

string

Platform URL. --base-url takes precedence; otherwise this overrides the saved URL. The public default is https://platform.toloka.ai.

string

Project ID. --project-id takes precedence; otherwise this overrides toloka.json and the saved project selection.

string

Credentials JSON file. Defaults to $XDG_CONFIG_HOME/toloka/credentials.json, or ~/.config/toloka/credentials.json.

string

Browser-session directory. Defaults to ~/.toloka/cli/credentials, or browser-sessions beside TOLOKA_CREDENTIALS_FILE when that file is set.

string

Base directory for the default credentials file. Defaults to ~/.config.

string

Home directory used for credential and browser-session paths. Defaults to the operating system’s home directory.

string

Executable used to open the sign-in URL. The URL is passed as its argument. Defaults to the operating system’s browser launcher.

string

Set to 0 to disable usage and exception reports, or 1 to enable them. An explicit value takes precedence over DO_NOT_TRACK and CI.

string

Set to 1 to disable usage and exception reports unless TOLOKA_TELEMETRY explicitly enables them.

string

Set to 1 to disable usage reports and automatic updates by default. TOLOKA_TELEMETRY and TOLOKA_AUTO_UPDATE can explicitly enable them.

string

Set to 1 to skip automatic updates, even when TOLOKA_AUTO_UPDATE enables them.

string

Set to 0 to disable automatic updates, or 1 to enable them. Otherwise the auto-update machine setting applies outside CI. Release eligibility and command exclusions still apply.

string

Any set value disables terminal colors unless FORCE_COLOR overrides it.

string

The value dumb disables terminal colors unless FORCE_COLOR overrides it.

string

Set to 0 to disable colors, or any other value to enable them even when output is piped. Overrides NO_COLOR and TERM.

string

Bun installation directory used by version reporting to identify the install channel. Defaults to ~/.bun.

string

Executable search path, also forwarded to local check processes.

string

Deprecated alias of TOLOKA_API_KEY. The current name takes precedence; using the alias prints a deprecation warning.

string

Deprecated alias of TOLOKA_BASE_URL. The current name takes precedence; using the alias prints a deprecation warning.

string

Deprecated alias of TOLOKA_PROJECT_ID. The current name takes precedence; using the alias prints a deprecation warning.

Exit codes

Code	Meaning
0	Command succeeded, or help/version was printed. A successful run start without `--wait` does not mean the run has finished. Successful `training status` and `batch-inference status` calls return 0 even for failed runs, including `batch-inference status --wait`; inspect the returned `status`.
1	Command or argument error, failed pipeline run or lint, or failed/missing local check.
2	Waiting for a pipeline run or its lints timed out, or a local check could not complete because its process or tooling failed.
130	SIGINT interrupted a non-watch check or the check stage of apply.
143	SIGTERM interrupted a non-watch check or the check stage of apply.

Reference

TypeScript SDK

CLI

HTTP API

Help

Commands

Global options

Additional help

Environment variables

Exit codes

​Help

​Commands

​Global options

​Additional help

​Environment variables

​Exit codes

Help

Commands

Global options

Additional help

Environment variables

Exit codes