Page agent

page.agent runs Skyvern’s agents against the page you already have open, then returns control to your script on the same page.

page.agent.run_task

Run a task in the context of this page and wait for it to finish.

await page.agent.run_task(
    prompt: str,
    engine: RunEngine = RunEngine.skyvern_v1,
    model: dict[str, Any] | None = None,
    url: str | None = None,
    webhook_url: str | None = None,
    totp_identifier: str | None = None,
    totp_url: str | None = None,
    title: str | None = None,
    error_code_mapping: dict[str, str] | None = None,
    data_extraction_schema: dict[str, Any] | str | None = None,
    max_steps: int | None = None,
    timeout: float = 1800,
    user_agent: str | None = None,
) -> TaskRunResponse

page.agent.runTask(prompt: string, options?: { engine?: RunEngine; model?: Record<string, unknown>; url?: string; webhookUrl?: string; totpIdentifier?: string; totpUrl?: string; title?: string; errorCodeMapping?: Record<string, string>; dataExtractionSchema?: Record<string, unknown> | string; maxSteps?: number; timeout?: number; }): Promise<TaskRunResponse>

Parameters

str

required

Natural language description of the task to perform.

RunEngine

The engine that runs the task. Defaults to skyvern_v1 ("skyvern-1.0"), suited to simple, single-goal tasks. skyvern_v3 ("skyvern-3.0") is the newest engine; the REST API and the Python SDK accept it, but the published TypeScript client’s type does not list it yet. skyvern_v2 remains for existing 2.0 runs. openai_cua and anthropic_cua use OpenAI’s and Anthropic’s computer-use models; ui_tars and yutori_navigator are also accepted.Default: RunEngine.skyvern_v1

dict[str, Any] | None

LLM model configuration options.Default: None

str | None

URL to navigate to. If not provided, uses the current page URL.Default: None

str | None

URL to receive webhook notifications about task progress.Default: None · TypeScript: webhookUrl

str | None

Identifier for TOTP (Time-based One-Time Password) authentication.Default: None · TypeScript: totpIdentifier

str | None

URL to fetch TOTP codes from.Default: None · TypeScript: totpUrl

str | None

Human-readable title for this task run.Default: None

dict[str, str] | None

Mapping of error codes to custom error messages.Default: None · TypeScript: errorCodeMapping

dict[str, Any] | str | None

Schema defining what data to extract from the page.Default: None · TypeScript: dataExtractionSchema

int | None

Maximum number of steps the agent can take.Default: None · TypeScript: maxSteps

float

Maximum time in seconds to wait for task completion.Default: 1800

str | None

Custom user agent string to use.Default: None · Python only

Returns TaskRunResponse · TypeScript Promise<TaskRunResponse> — TaskRunResponse containing the task execution results.

TaskRunResponse fields

Field	Type	Description
`run_id`	`str`	Unique identifier for this run. Run ID starts with `tsk_` for task runs and `wr_` for workflow runs.
`status`	`"created" \| "queued" \| "running" \| "paused" \| "timed_out" \| "failed" \| "terminated" \| "completed" \| "canceled"`	Current status of the run
`output`	`dict[str, Any \| None] \| list[Any \| None] \| str \| None`	Output data from the run, if any. Format/schema depends on the data extracted by the run.
`downloaded_files`	`list[FileInfo] \| None`	List of files downloaded during the run
`recording_url`	`str \| None`	URL to the recording of the run
`recording_archived`	`bool \| None`	True when the recording exists but has been archived to cold storage and is not currently accessible.
`screenshot_urls`	`list[str] \| None`	List of last n screenshot URLs in reverse chronological order - the first one the list is the latest screenshot.
`failure_reason`	`str \| None`	Reason for failure if the run failed or terminated
`created_at`	`datetime`	Timestamp when this run was created
`modified_at`	`datetime`	Timestamp when this run was last modified
`queued_at`	`datetime \| None`	Timestamp when this run was queued
`started_at`	`datetime \| None`	Timestamp when this run started execution
`finished_at`	`datetime \| None`	Timestamp when this run finished
`app_url`	`str \| None`	URL to the application UI where the run can be viewed
`browser_session_id`	`str \| None`	ID of the Skyvern persistent browser session used for this run
`browser_profile_id`	`str \| None`	ID of the browser profile used for this run
`max_screenshot_scrolls`	`int \| None`	The maximum number of scrolls for the post action screenshot. When it’s None or 0, it takes the current viewpoint screenshot
`script_run`	`ScriptRunResponse \| None`	The script run result
`errors`	`list[dict[str, Any \| None]] \| None`	The errors for the run
`step_count`	`int \| None`	Total number of steps executed in this run
`run_request`	`TaskRunRequestOutput \| None`	The original request parameters used to start this task run

Example

import asyncio

from skyvern import Skyvern


async def main() -> None:
    skyvern = Skyvern(api_key="YOUR_API_KEY")
    browser = await skyvern.launch_cloud_browser()
    page = await browser.get_working_page()
    await page.goto("https://example.com")
    result = await page.agent.run_task(prompt="Find the top post and return its title")
    print(result)
    await browser.close()


asyncio.run(main())

import { Skyvern } from "@skyvern/client";

const skyvern = new Skyvern({ apiKey: "YOUR_API_KEY" });
const browser = await skyvern.launchCloudBrowser();
const page = await browser.getWorkingPage();
await page.goto("https://example.com");
const result = await page.agent.runTask("Find the top post and return its title");
console.log(result);
await browser.close();

Run a login task in the context of this page and wait for it to finish. This method has multiple overloaded signatures for different credential types:

Skyvern credentials (default):

await page.agent.login(credential_id="cred_123")

Bitwarden credentials:

await page.agent.login(
    credential_type=CredentialType.bitwarden,
    bitwarden_collection_id="collection_id",
    bitwarden_item_id="item_id",
)

1Password credentials:

await page.agent.login(
    credential_type=CredentialType.onepassword,
    onepassword_vault_id="vault_id",
    onepassword_item_id="item_id",
)

Azure Vault credentials:

await page.agent.login(
    credential_type=CredentialType.azure_vault,
    azure_vault_name="vault_name",
    azure_vault_username_key="username_key",
    azure_vault_password_key="password_key",
)

await page.agent.login(
    *,
    credential_type: CredentialType | str = CredentialType.skyvern,
    url: str | None = None,
    credential_id: str | None = None,
    bitwarden_collection_id: str | None = None,
    bitwarden_item_id: str | None = None,
    onepassword_vault_id: str | None = None,
    onepassword_item_id: str | None = None,
    azure_vault_name: str | None = None,
    azure_vault_username_key: str | None = None,
    azure_vault_password_key: str | None = None,
    azure_vault_totp_secret_key: str | None = None,
    prompt: str | None = None,
    webhook_url: str | None = None,
    totp_identifier: str | None = None,
    totp_url: str | None = None,
    extra_http_headers: dict[str, str] | None = None,
    timeout: float = 1800,
) -> WorkflowRunResponse

page.agent.login(credentialType: "skyvern", options: { credentialId: string; url?: string; prompt?: string; webhookUrl?: string; totpIdentifier?: string; totpUrl?: string; extraHttpHeaders?: Record<string, string>; timeout?: number; }): Promise<WorkflowRunResponse>
page.agent.login(credentialType: "bitwarden", options: { bitwardenItemId: string; bitwardenCollectionId?: string; url?: string; prompt?: string; webhookUrl?: string; totpIdentifier?: string; totpUrl?: string; extraHttpHeaders?: Record<string, string>; timeout?: number; }): Promise<WorkflowRunResponse>
page.agent.login(credentialType: "1password", options: { onepasswordVaultId: string; onepasswordItemId: string; url?: string; prompt?: string; webhookUrl?: string; totpIdentifier?: string; totpUrl?: string; extraHttpHeaders?: Record<string, string>; timeout?: number; }): Promise<WorkflowRunResponse>
page.agent.login(credentialType: "azure_vault", options: { azureVaultName: string; azureVaultUsernameKey: string; azureVaultPasswordKey: string; azureVaultTotpSecretKey?: string; url?: string; prompt?: string; webhookUrl?: string; totpIdentifier?: string; totpUrl?: string; extraHttpHeaders?: Record<string, string>; timeout?: number; }): Promise<WorkflowRunResponse>

Parameters

CredentialType | str

Type of credential store to use. Defaults to CredentialType.skyvern.Default: CredentialType.skyvern · TypeScript: credentialType (required)

str | None

URL to navigate to for login. If not provided, uses the current page URL.Default: None

str | None

ID of the Skyvern credential to use (required for skyvern credential_type).Default: None · TypeScript: credentialId

str | None

Bitwarden collection ID (optional for bitwarden credential_type).Default: None · TypeScript: bitwardenCollectionId

str | None

Bitwarden item ID (required for bitwarden credential_type).Default: None · TypeScript: bitwardenItemId

str | None

1Password vault ID (required for onepassword credential_type).Default: None · TypeScript: onepasswordVaultId

str | None

1Password item ID (required for onepassword credential_type).Default: None · TypeScript: onepasswordItemId

str | None

Azure Vault name (required for azure_vault credential_type).Default: None · TypeScript: azureVaultName

str | None

Azure Vault username key (required for azure_vault credential_type).Default: None · TypeScript: azureVaultUsernameKey

str | None

Azure Vault password key (required for azure_vault credential_type).Default: None · TypeScript: azureVaultPasswordKey

str | None

Azure Vault TOTP secret key (optional for azure_vault credential_type).Default: None · TypeScript: azureVaultTotpSecretKey

str | None

Additional instructions for the login process.Default: None

str | None

URL to receive webhook notifications about login progress.Default: None · TypeScript: webhookUrl

str | None

Identifier for TOTP authentication.Default: None · TypeScript: totpIdentifier

str | None

URL to fetch TOTP codes from.Default: None · TypeScript: totpUrl

dict[str, str] | None

Additional HTTP headers to include in requests.Default: None · TypeScript: extraHttpHeaders

float

Maximum time in seconds to wait for login completion.Default: 1800

Returns WorkflowRunResponse · TypeScript Promise<WorkflowRunResponse> — WorkflowRunResponse containing the login workflow execution results.

WorkflowRunResponse fields

Field	Type	Description
`run_id`	`str`	Unique identifier for this run. Run ID starts with `tsk_` for task runs and `wr_` for workflow runs.
`status`	`"created" \| "queued" \| "running" \| "paused" \| "timed_out" \| "failed" \| "terminated" \| "completed" \| "canceled"`	Current status of the run
`output`	`dict[str, Any \| None] \| list[Any \| None] \| str \| None`	Output data from the run, if any. Format/schema depends on the data extracted by the run.
`downloaded_files`	`list[FileInfo] \| None`	List of files downloaded during the run
`recording_url`	`str \| None`	URL to the recording of the run
`recording_archived`	`bool \| None`	True when the recording exists but has been archived to cold storage and is not currently accessible.
`screenshot_urls`	`list[str] \| None`	List of last n screenshot URLs in reverse chronological order - the first one the list is the latest screenshot.
`failure_reason`	`str \| None`	Reason for failure if the run failed or terminated
`created_at`	`datetime`	Timestamp when this run was created
`modified_at`	`datetime`	Timestamp when this run was last modified
`queued_at`	`datetime \| None`	Timestamp when this run was queued
`started_at`	`datetime \| None`	Timestamp when this run started execution
`finished_at`	`datetime \| None`	Timestamp when this run finished
`app_url`	`str \| None`	URL to the application UI where the run can be viewed
`browser_session_id`	`str \| None`	ID of the Skyvern persistent browser session used for this run
`browser_profile_id`	`str \| None`	ID of the browser profile used for this run
`max_screenshot_scrolls`	`int \| None`	The maximum number of scrolls for the post action screenshot. When it’s None or 0, it takes the current viewpoint screenshot
`script_run`	`ScriptRunResponse \| None`	The script run result
`errors`	`list[dict[str, Any \| None]] \| None`	The errors for the run
`step_count`	`int \| None`	Total number of steps executed in this run
`run_with`	`str \| None`	Whether the workflow run was executed with agent or code
`ai_fallback`	`bool \| None`	Whether to fallback to AI if code run fails.
`script_id`	`str \| None`	ID of the cached script used for this workflow run, if any.
`run_request`	`WorkflowRunRequestOutput \| None`	The original request parameters used to start this workflow run

Example

import asyncio

from skyvern import Skyvern


async def main() -> None:
    skyvern = Skyvern(api_key="YOUR_API_KEY")
    browser = await skyvern.launch_cloud_browser()
    page = await browser.get_working_page()
    await page.goto("https://example.com")
    result = await page.agent.login(credential_id="cred_123", url="https://example.com/login")
    print(result)
    await browser.close()


asyncio.run(main())

import { Skyvern } from "@skyvern/client";

const skyvern = new Skyvern({ apiKey: "YOUR_API_KEY" });
const browser = await skyvern.launchCloudBrowser();
const page = await browser.getWorkingPage();
await page.goto("https://example.com");
const result = await page.agent.login("skyvern", {
    credentialId: "cred_123",
    url: "https://example.com/login",
});
console.log(result);
await browser.close();

page.agent.download_files

Run a file download task in the context of this page and wait for it to finish.

await page.agent.download_files(
    prompt: str,
    *,
    url: str | None = None,
    download_suffix: str | None = None,
    download_timeout: float | None = None,
    max_steps_per_run: int | None = None,
    webhook_url: str | None = None,
    totp_identifier: str | None = None,
    totp_url: str | None = None,
    extra_http_headers: dict[str, str] | None = None,
    timeout: float = 1800,
) -> WorkflowRunResponse

page.agent.downloadFiles(prompt: string, options?: { url?: string; downloadSuffix?: string; downloadTimeout?: number; maxStepsPerRun?: number; webhookUrl?: string; totpIdentifier?: string; totpUrl?: string; extraHttpHeaders?: Record<string, string>; timeout?: number; }): Promise<WorkflowRunResponse>

Parameters

str

required

Instructions for navigating to and downloading the file.

str | None

URL to navigate to for file download. If not provided, uses the current page URL.Default: None

str | None

Complete filename for the downloaded file. Supports {{ original_filename }} to keep the name the site gave the file.Default: None · TypeScript: downloadSuffix

float | None

Timeout in seconds for the download operation.Default: None · TypeScript: downloadTimeout

int | None

Maximum number of steps to execute.Default: None · TypeScript: maxStepsPerRun

str | None

URL to receive webhook notifications about download progress.Default: None · TypeScript: webhookUrl

str | None

Identifier for TOTP authentication.Default: None · TypeScript: totpIdentifier

str | None

URL to fetch TOTP codes from.Default: None · TypeScript: totpUrl

dict[str, str] | None

Additional HTTP headers to include in requests.Default: None · TypeScript: extraHttpHeaders

float

Maximum time in seconds to wait for download completion.Default: 1800

Returns WorkflowRunResponse · TypeScript Promise<WorkflowRunResponse> — WorkflowRunResponse containing the file download workflow execution results.

WorkflowRunResponse fields

Field	Type	Description
`run_id`	`str`	Unique identifier for this run. Run ID starts with `tsk_` for task runs and `wr_` for workflow runs.
`status`	`"created" \| "queued" \| "running" \| "paused" \| "timed_out" \| "failed" \| "terminated" \| "completed" \| "canceled"`	Current status of the run
`output`	`dict[str, Any \| None] \| list[Any \| None] \| str \| None`	Output data from the run, if any. Format/schema depends on the data extracted by the run.
`downloaded_files`	`list[FileInfo] \| None`	List of files downloaded during the run
`recording_url`	`str \| None`	URL to the recording of the run
`recording_archived`	`bool \| None`	True when the recording exists but has been archived to cold storage and is not currently accessible.
`screenshot_urls`	`list[str] \| None`	List of last n screenshot URLs in reverse chronological order - the first one the list is the latest screenshot.
`failure_reason`	`str \| None`	Reason for failure if the run failed or terminated
`created_at`	`datetime`	Timestamp when this run was created
`modified_at`	`datetime`	Timestamp when this run was last modified
`queued_at`	`datetime \| None`	Timestamp when this run was queued
`started_at`	`datetime \| None`	Timestamp when this run started execution
`finished_at`	`datetime \| None`	Timestamp when this run finished
`app_url`	`str \| None`	URL to the application UI where the run can be viewed
`browser_session_id`	`str \| None`	ID of the Skyvern persistent browser session used for this run
`browser_profile_id`	`str \| None`	ID of the browser profile used for this run
`max_screenshot_scrolls`	`int \| None`	The maximum number of scrolls for the post action screenshot. When it’s None or 0, it takes the current viewpoint screenshot
`script_run`	`ScriptRunResponse \| None`	The script run result
`errors`	`list[dict[str, Any \| None]] \| None`	The errors for the run
`step_count`	`int \| None`	Total number of steps executed in this run
`run_with`	`str \| None`	Whether the workflow run was executed with agent or code
`ai_fallback`	`bool \| None`	Whether to fallback to AI if code run fails.
`script_id`	`str \| None`	ID of the cached script used for this workflow run, if any.
`run_request`	`WorkflowRunRequestOutput \| None`	The original request parameters used to start this workflow run

Example

import asyncio

from skyvern import Skyvern


async def main() -> None:
    skyvern = Skyvern(api_key="YOUR_API_KEY")
    browser = await skyvern.launch_cloud_browser()
    page = await browser.get_working_page()
    await page.goto("https://example.com")
    result = await page.agent.download_files(prompt="Download the latest invoice")
    print(result)
    await browser.close()


asyncio.run(main())

import { Skyvern } from "@skyvern/client";

const skyvern = new Skyvern({ apiKey: "YOUR_API_KEY" });
const browser = await skyvern.launchCloudBrowser();
const page = await browser.getWorkingPage();
await page.goto("https://example.com");
const result = await page.agent.downloadFiles("Download the latest invoice");
console.log(result);
await browser.close();

page.agent.run_workflow

Run a workflow in the context of this page and wait for it to finish.

await page.agent.run_workflow(
    workflow_id: str,
    parameters: dict[str, Any] | None = None,
    template: bool | None = None,
    title: str | None = None,
    webhook_url: str | None = None,
    totp_url: str | None = None,
    totp_identifier: str | None = None,
    max_elapsed_time_minutes: int | None = None,
    run_metadata: dict[str, str] | None = None,
    timeout: float = 1800,
) -> WorkflowRunResponse

page.agent.runWorkflow(workflowId: string, options?: { parameters?: Record<string, unknown>; template?: boolean; title?: string; webhookUrl?: string; totpUrl?: string; totpIdentifier?: string; timeout?: number; }): Promise<WorkflowRunResponse>

Parameters

str

required

ID of the workflow to execute.TypeScript: workflowId

dict[str, Any] | None

Dictionary of parameters to pass to the workflow.Default: None

bool | None

Whether this is a workflow template.Default: None

str | None

Human-readable title for this workflow run.Default: None

str | None

URL to receive webhook notifications about workflow progress.Default: None · TypeScript: webhookUrl

str | None

URL to fetch TOTP codes from.Default: None · TypeScript: totpUrl

str | None

Identifier for TOTP authentication.Default: None · TypeScript: totpIdentifier

int | None

Timeout this workflow run after the configured elapsed runtime in minutes.Default: None · Python only

dict[str, str] | None

String key/value metadata to attach to this workflow run.Default: None · Python only

float

Maximum time in seconds to wait for workflow completion.Default: 1800

Returns WorkflowRunResponse · TypeScript Promise<WorkflowRunResponse> — WorkflowRunResponse containing the workflow execution results.

WorkflowRunResponse fields

Field	Type	Description
`run_id`	`str`	Unique identifier for this run. Run ID starts with `tsk_` for task runs and `wr_` for workflow runs.
`status`	`"created" \| "queued" \| "running" \| "paused" \| "timed_out" \| "failed" \| "terminated" \| "completed" \| "canceled"`	Current status of the run
`output`	`dict[str, Any \| None] \| list[Any \| None] \| str \| None`	Output data from the run, if any. Format/schema depends on the data extracted by the run.
`downloaded_files`	`list[FileInfo] \| None`	List of files downloaded during the run
`recording_url`	`str \| None`	URL to the recording of the run
`recording_archived`	`bool \| None`	True when the recording exists but has been archived to cold storage and is not currently accessible.
`screenshot_urls`	`list[str] \| None`	List of last n screenshot URLs in reverse chronological order - the first one the list is the latest screenshot.
`failure_reason`	`str \| None`	Reason for failure if the run failed or terminated
`created_at`	`datetime`	Timestamp when this run was created
`modified_at`	`datetime`	Timestamp when this run was last modified
`queued_at`	`datetime \| None`	Timestamp when this run was queued
`started_at`	`datetime \| None`	Timestamp when this run started execution
`finished_at`	`datetime \| None`	Timestamp when this run finished
`app_url`	`str \| None`	URL to the application UI where the run can be viewed
`browser_session_id`	`str \| None`	ID of the Skyvern persistent browser session used for this run
`browser_profile_id`	`str \| None`	ID of the browser profile used for this run
`max_screenshot_scrolls`	`int \| None`	The maximum number of scrolls for the post action screenshot. When it’s None or 0, it takes the current viewpoint screenshot
`script_run`	`ScriptRunResponse \| None`	The script run result
`errors`	`list[dict[str, Any \| None]] \| None`	The errors for the run
`step_count`	`int \| None`	Total number of steps executed in this run
`run_with`	`str \| None`	Whether the workflow run was executed with agent or code
`ai_fallback`	`bool \| None`	Whether to fallback to AI if code run fails.
`script_id`	`str \| None`	ID of the cached script used for this workflow run, if any.
`run_request`	`WorkflowRunRequestOutput \| None`	The original request parameters used to start this workflow run

Example

import asyncio

from skyvern import Skyvern


async def main() -> None:
    skyvern = Skyvern(api_key="YOUR_API_KEY")
    browser = await skyvern.launch_cloud_browser()
    page = await browser.get_working_page()
    await page.goto("https://example.com")
    result = await page.agent.run_workflow(workflow_id="wpid_123")
    print(result)
    await browser.close()


asyncio.run(main())

import { Skyvern } from "@skyvern/client";

const skyvern = new Skyvern({ apiKey: "YOUR_API_KEY" });
const browser = await skyvern.launchCloudBrowser();
const page = await browser.getWorkingPage();
await page.goto("https://example.com");
const result = await page.agent.runWorkflow("wpid_123");
console.log(result);
await browser.close();

SDK

Browser automation

API client

page.agent.run_task

page.agent.download_files

page.agent.run_workflow

​page.agent.run_task

​page.agent.login

​page.agent.download_files

​page.agent.run_workflow

page.agent.run_task

page.agent.login

page.agent.download_files

page.agent.run_workflow