同步完整源码 - 2026-05-24
This commit is contained in:
@@ -0,0 +1,10 @@
|
||||
---
|
||||
title: AI Agents Quickstart
|
||||
subtitle: Give your AI coding assistant full browser automation in one command
|
||||
description: Set up Skyvern's MCP server so Claude, Cursor, Windsurf, or any AI coding assistant can automate browsers with a single command.
|
||||
slug: cloud/getting-started/ai-agents-quickstart
|
||||
---
|
||||
|
||||
import AIAgentsQuickstartContent from "/snippets/ai-agents-quickstart-content.mdx";
|
||||
|
||||
<AIAgentsQuickstartContent />
|
||||
@@ -0,0 +1,247 @@
|
||||
---
|
||||
title: Core Concepts
|
||||
subtitle: The building blocks of every automation in the Skyvern dashboard
|
||||
description: Understand tasks, workflows, blocks, runs, browser sessions, browser profiles, and credentials in the Skyvern Cloud UI — what each is, where to find it, and how it fits together.
|
||||
slug: cloud/getting-started/core-concepts
|
||||
keywords:
|
||||
- task
|
||||
- workflow
|
||||
- block
|
||||
- browser session
|
||||
- browser profile
|
||||
- credential
|
||||
- run
|
||||
- schedule
|
||||
- engine
|
||||
---
|
||||
|
||||
Every automation you build in Skyvern Cloud is composed of the same nine building blocks. This page is a quick tour of each one — what it is, where you find it in the dashboard, and when to reach for it.
|
||||
|
||||
If you've already poked around the dashboard, this map will tie what you've seen to the underlying concepts.
|
||||
|
||||
---
|
||||
|
||||
## Tasks
|
||||
|
||||
A **Task** is a single automation job. You describe what you want, Skyvern opens a browser and does it.
|
||||
|
||||
You create tasks from the **Discover** page. Type your prompt and target URL into the input bar, pick an engine, and hit send.
|
||||
|
||||
<img src="/images/cloud/discover-page-overview.png" alt="Discover page where you create one-off tasks" />
|
||||
|
||||
Tasks are the right starting point for one-off automations, ad-hoc data extraction, or quickly testing whether an idea is feasible before turning it into a workflow.
|
||||
|
||||
When you outgrow a single task — you need parameters, branching, loops, or to run the same flow on a schedule — graduate to a Workflow.
|
||||
|
||||
<Card title="Run a task →" icon="play" href="/cloud/getting-started/run-a-task">
|
||||
Walkthrough of every option on the Discover page.
|
||||
</Card>
|
||||
|
||||
---
|
||||
|
||||
## Workflows
|
||||
|
||||
A **Workflow** is a reusable automation built by chaining blocks together on a visual canvas. Workflows are versioned, parameterized, and shareable across your team.
|
||||
|
||||
You build them on the **Workflows** page. Click any workflow to open the editor, where blocks appear as connected nodes you can drag and configure.
|
||||
|
||||
<img src="/images/cloud/workflow-editor-overview.png" alt="Workflow editor with blocks chained together" />
|
||||
|
||||
Use a workflow when:
|
||||
|
||||
- The same flow needs to run repeatedly with different inputs
|
||||
- You need to loop over a list, branch on a condition, or pass data between steps
|
||||
- You want to schedule the automation to run on a cadence
|
||||
- You want a teammate to be able to re-run it without re-deriving the logic
|
||||
|
||||
<Card title="Build a workflow →" icon="diagram-project" href="/cloud/building-workflows/build-a-workflow">
|
||||
Tutorial covering blocks, parameters, branching, and execution.
|
||||
</Card>
|
||||
|
||||
---
|
||||
|
||||
## Blocks
|
||||
|
||||
**Blocks** are the units a workflow is built from. Each block performs one thing — navigate, log in, extract data, branch on a condition, send an email — and passes its output to the next block.
|
||||
|
||||
In the workflow editor, click the **+** button on any node to insert a new block. The picker groups every available block by category.
|
||||
|
||||
<img src="/images/cloud/add-block-panel.png" alt="Block picker with categories and block types" />
|
||||
|
||||
There are 20+ block types organized into:
|
||||
|
||||
- **Browser automation** — Browser Task, Browser Action, Login, Extraction, Go to URL, Print Page
|
||||
- **Data and extraction** — Text Prompt, File Parser
|
||||
- **Control flow** — Loop, Conditional, AI Validation, Code, Wait
|
||||
- **Files** — File Download, Cloud Storage Upload
|
||||
- **Communication** — Send Email, HTTP Request, Human Interaction
|
||||
|
||||
Each block has its own configuration panel. The full reference, with every field for every block, lives at [Workflow Blocks](/cloud/building-workflows/configure-blocks).
|
||||
|
||||
---
|
||||
|
||||
## Runs
|
||||
|
||||
Every time a task or workflow executes, Skyvern creates a **Run** — a record of what happened, with a video, screenshots, action log, and any extracted data.
|
||||
|
||||
The **Runs** page is your history. Filter by status, search by run ID, click any row to drill in.
|
||||
|
||||
<img src="/images/cloud/run-history-overview.png" alt="Runs page showing execution history with statuses" />
|
||||
|
||||
Inside a run detail page you'll find:
|
||||
|
||||
- **Live viewer / recording** — full video of the browser as it executed
|
||||
- **Actions** — every step the agent took, with the LLM's reasoning
|
||||
- **Output** — the structured data extracted (matches the schema you defined)
|
||||
- **Code** — auto-generated Python/TypeScript/cURL to reproduce this run via API
|
||||
- **Downloaded files** — anything Skyvern saved, with signed URLs and checksums
|
||||
|
||||
A run can be `queued`, `running`, `completed`, `failed`, `terminated`, `timed_out`, or `canceled`. You're billed per **step** — one screenshot + LLM decision + browser action.
|
||||
|
||||
<Card title="Read run details →" icon="list-check" href="/cloud/viewing-results/run-details">
|
||||
How to interpret the actions tab, recordings, and output.
|
||||
</Card>
|
||||
|
||||
---
|
||||
|
||||
## Browser sessions
|
||||
|
||||
A **browser session** is a live browser instance you keep open across multiple runs. Sessions preserve everything that lives in the browser — cookies, login state, cached pages, the current tab — so a follow-up task picks up exactly where the previous one left off.
|
||||
|
||||
You manage sessions on the **Browsers** page. Create a session, run tasks against it, then close it when you're done.
|
||||
|
||||
<img src="/images/cloud/browsers-page-overview.png" alt="Browsers page with active session and details" />
|
||||
|
||||
Sessions are perfect when you want to:
|
||||
|
||||
- Log in once, then run a sequence of tasks behind that login
|
||||
- Hand off a partially-completed flow between an agent and a human (Take Control)
|
||||
- Keep a long-running automation warm without re-authenticating each step
|
||||
|
||||
Sessions live for 5 minutes to 24 hours (default 60 minutes). When the timer ends or you close it manually, the browser is destroyed and its state goes with it.
|
||||
|
||||
<Card title="Browser sessions →" icon="window" href="/cloud/browser-management/browser-sessions">
|
||||
Create, attach, and reuse sessions across runs.
|
||||
</Card>
|
||||
|
||||
---
|
||||
|
||||
## Browser profiles
|
||||
|
||||
A **browser profile** is the persistent counterpart to a session. Where a session is a live browser that times out, a profile is a saved snapshot — cookies, auth tokens, local storage — that you can reload weeks later to skip the login flow entirely.
|
||||
|
||||
Profiles are listed on the **Browsers** page under the Profiles tab. You create one from any completed run, then attach it to future runs to start in an already-authenticated state.
|
||||
|
||||
| | Browser session | Browser profile |
|
||||
|---|---|---|
|
||||
| **Live or saved?** | Live, in-memory browser | Saved snapshot of state |
|
||||
| **Lifetime** | 5 min – 24 hr | Indefinite |
|
||||
| **Best for** | Chaining tasks in real time | Skipping login on repeat runs |
|
||||
|
||||
<Card title="Browser profiles →" icon="floppy-disk" href="/cloud/browser-management/browser-profiles">
|
||||
Save and reuse authenticated state across days or weeks.
|
||||
</Card>
|
||||
|
||||
---
|
||||
|
||||
## Credentials
|
||||
|
||||
A **credential** is a stored login, credit card, or 2FA secret that Skyvern injects directly into the browser at runtime. Credentials are encrypted at rest and **never sent to the LLM** — the agent decides where the password field is, and the value is filled in via the browser layer.
|
||||
|
||||
The **Credentials** page is where you manage them.
|
||||
|
||||
<img src="/images/cloud/credentials-overview.png" alt="Credentials page with password and credit card credentials" />
|
||||
|
||||
Three credential types are supported:
|
||||
|
||||
- **Password** — username + password, optionally with a TOTP secret or email/SMS 2FA
|
||||
- **Credit card** — card number, expiry, CVC, billing address
|
||||
- **Secret** — any single value you want to keep out of the LLM context
|
||||
|
||||
You can also sync credentials from **Bitwarden**, **1Password**, or **Azure Key Vault** instead of storing them in Skyvern.
|
||||
|
||||
<Card title="Credentials overview →" icon="lock" href="/cloud/managing-credentials/credentials-overview">
|
||||
Security model, supported types, and external vault integration.
|
||||
</Card>
|
||||
|
||||
---
|
||||
|
||||
## Schedules
|
||||
|
||||
A **schedule** runs a workflow automatically on a recurring cadence. You set a cron expression and timezone, and Skyvern fires the workflow at every interval — no manual trigger needed.
|
||||
|
||||
Schedules are configured per workflow. Open any workflow's actions menu and choose **Schedule**.
|
||||
|
||||
<Frame>
|
||||
<img src="/images/cloud/create-sched-form.png" alt="Schedule form with cron expression and timezone" width="300" />
|
||||
</Frame>
|
||||
|
||||
Use schedules for daily reports, hourly monitoring, periodic data syncs, or anything that should "just keep running" without you remembering to click Run.
|
||||
|
||||
<Card title="Scheduling →" icon="calendar" href="/cloud/building-workflows/scheduling">
|
||||
Cron syntax, timezones, and managing scheduled runs.
|
||||
</Card>
|
||||
|
||||
---
|
||||
|
||||
## Engines
|
||||
|
||||
An **engine** is the AI model Skyvern uses to drive the browser. Different engines have different strengths — speed, accuracy on complex pages, cost — and you can pick one per task or per block.
|
||||
|
||||
The engine picker appears on the Discover page (for tasks) and on every browser-driven block in the workflow editor.
|
||||
|
||||
| Engine | When to use |
|
||||
|--------|-------------|
|
||||
| **Skyvern 2.0** | Default. Best balance of accuracy and speed for most automations. |
|
||||
| **Skyvern 1.0** | Lighter and faster for simple, single-page interactions. |
|
||||
| **OpenAI CUA** | OpenAI's Computer Use Agent. Good for visual reasoning. |
|
||||
| **Anthropic CUA** | Anthropic's Claude Computer Use. Strong at multi-step planning. |
|
||||
| **UI-TARS** | Lightweight CUA model option. |
|
||||
|
||||
If you're not sure, leave it on the default. Switch engines if you hit accuracy issues on a specific site.
|
||||
|
||||
---
|
||||
|
||||
## How they fit together
|
||||
|
||||
```mermaid
|
||||
flowchart LR
|
||||
A[Task] -->|Single job, one prompt| R1[Run]
|
||||
W[Workflow] -->|Reusable, multi-step| R2[Run]
|
||||
W -->|Composed of| B[Blocks]
|
||||
S[Schedule] -->|Triggers| W
|
||||
R1 & R2 -->|Optionally use| BS[Browser Session]
|
||||
BS -->|Snapshot to| BP[Browser Profile]
|
||||
BP -->|Restored by| R1
|
||||
BP -->|Restored by| R2
|
||||
R1 & R2 -->|Inject at runtime| C[Credentials]
|
||||
R1 & R2 -->|Driven by| E[Engine]
|
||||
```
|
||||
|
||||
A typical day in the dashboard:
|
||||
|
||||
1. **Prototype** with a Task on the Discover page. Iterate on the prompt until it works.
|
||||
2. **Productionize** by turning the prompt into a Workflow with Blocks, Parameters, and a Schedule.
|
||||
3. **Persist state** by saving a Browser Profile after the first authenticated run, so future runs skip login.
|
||||
4. **Audit** every execution on the Runs page — recording, actions, output, downloaded files.
|
||||
|
||||
---
|
||||
|
||||
## Next steps
|
||||
|
||||
<CardGroup cols={2}>
|
||||
<Card
|
||||
title="Run Your First Task"
|
||||
icon="play"
|
||||
href="/cloud/getting-started/run-your-first-task"
|
||||
>
|
||||
Hands-on tutorial — go from prompt to extracted data in a few minutes.
|
||||
</Card>
|
||||
<Card
|
||||
title="Build a Workflow"
|
||||
icon="diagram-project"
|
||||
href="/cloud/building-workflows/build-a-workflow"
|
||||
>
|
||||
Wire blocks together into a reusable, parameterized automation.
|
||||
</Card>
|
||||
</CardGroup>
|
||||
@@ -0,0 +1,17 @@
|
||||
---
|
||||
title: What is Skyvern?
|
||||
subtitle: AI-powered browser automation for any website
|
||||
description: Learn what Skyvern is and how it uses AI and computer vision to automate browser interactions on any website through the Cloud UI.
|
||||
slug: cloud/getting-started/introduction
|
||||
keywords:
|
||||
- browser automation
|
||||
- AI agent
|
||||
- computer vision
|
||||
- LLM
|
||||
- natural language
|
||||
- visual reasoning
|
||||
---
|
||||
|
||||
import IntroductionContent from "/snippets/introduction-content.mdx";
|
||||
|
||||
<IntroductionContent />
|
||||
@@ -0,0 +1,25 @@
|
||||
---
|
||||
title: MCP Server
|
||||
subtitle: Connect AI assistants to browser automation via Model Context Protocol
|
||||
description: Install and configure Skyvern's MCP server so Claude Desktop, Claude Code, Cursor, Windsurf, Codex, Hermes, OpenClaw, and other AI tools can run browser automations from the Skyvern Cloud dashboard over OAuth or API key.
|
||||
slug: cloud/getting-started/mcp
|
||||
keywords:
|
||||
- MCP
|
||||
- Model Context Protocol
|
||||
- OAuth
|
||||
- Claude Desktop
|
||||
- Claude Code
|
||||
- Cursor
|
||||
- Windsurf
|
||||
- Codex
|
||||
- Hermes
|
||||
- OpenClaw
|
||||
- MCPorter
|
||||
- AI assistant
|
||||
- tools
|
||||
- streamable-http
|
||||
---
|
||||
|
||||
import McpContent from "/snippets/mcp-content.mdx";
|
||||
|
||||
<McpContent />
|
||||
@@ -0,0 +1,100 @@
|
||||
---
|
||||
title: Watching Live Execution
|
||||
subtitle: Monitor, interact with, and control running tasks
|
||||
description: Monitor running tasks in real time in the Skyvern Cloud UI, view the live browser stream and agent reasoning logs, and take manual control of the browser mid-task.
|
||||
slug: cloud/monitor-a-run
|
||||
keywords:
|
||||
- live viewer
|
||||
- real-time
|
||||
- agent reasoning
|
||||
- take control
|
||||
- manual control
|
||||
- step timeline
|
||||
---
|
||||
|
||||
When you run a task from the [Discover page](/cloud/getting-started/run-a-task), you're taken to the live execution screen where you can watch the browser in real time.
|
||||
|
||||
<img src="/images/cloud/live-execution-overview.png" alt="Live execution screen" />
|
||||
|
||||
---
|
||||
|
||||
## The execution screen
|
||||
|
||||
The execution view has three panels:
|
||||
|
||||
| Panel | What it shows |
|
||||
|-------|---------------|
|
||||
| **Left: Task configuration** | The block being executed, its URL, and prompt. A status badge shows the current state. |
|
||||
| **Center: Live browser** | Real-time view of the browser. You see pages load, forms fill, and buttons click. |
|
||||
| **Right: Agent logs** | Real-time LLM reasoning and action decisions. Shows why the AI made each choice. |
|
||||
|
||||
---
|
||||
|
||||
## When the live view is available
|
||||
|
||||
The live browser stream is active while the task is still in progress:
|
||||
|
||||
| Status | Live view |
|
||||
|--------|-----------|
|
||||
| `created` | Waiting to start |
|
||||
| `queued` | Waiting for a browser |
|
||||
| `running` | **Active**: the browser is navigating |
|
||||
| `completed` | Stream closed. View the recording instead. |
|
||||
| `failed` | Stream closed. View the recording instead. |
|
||||
| `terminated` | Stream closed. View the recording instead. |
|
||||
| `timed_out` | Stream closed. View the recording instead. |
|
||||
| `canceled` | Stream closed. View the recording instead. |
|
||||
|
||||
Once a task reaches a final state, the live stream closes. Open the run from **Runs** in the sidebar to access the full recording, screenshots, and action history.
|
||||
|
||||
---
|
||||
|
||||
## Taking control of the browser
|
||||
|
||||
While a task is running, you can take over the browser and interact with the page directly, just like you would in your own Chrome window.
|
||||
|
||||
### How it works
|
||||
|
||||
1. Click the **Take Control** button in the live browser panel.
|
||||
2. The AI agent pauses and you get full mouse and keyboard control of the browser.
|
||||
3. Interact with the page: type into fields, click buttons, solve CAPTCHAs, dismiss popups.
|
||||
4. Click **Release Control** to hand the browser back to the AI. It picks up from the current page state and continues the task.
|
||||
|
||||
### When to use it
|
||||
|
||||
- **CAPTCHAs**: solve a CAPTCHA the AI can't get past, then release control
|
||||
- **Unusual login flows**: complete a multi-factor or SSO login that the agent doesn't handle
|
||||
- **Unexpected popups or modals**: dismiss cookie banners, age gates, or promotional overlays
|
||||
- **Mid-task corrections**: navigate to the right page or fix a form field before letting the agent continue
|
||||
|
||||
<Warning>
|
||||
Taking control pauses the AI agent. The task timer keeps running, so release control promptly to avoid timeouts.
|
||||
</Warning>
|
||||
|
||||
<Tip>
|
||||
You can take and release control multiple times during a single run. Each time, the agent resumes from whatever state you left the browser in.
|
||||
</Tip>
|
||||
|
||||
---
|
||||
|
||||
## Stopping a running task
|
||||
|
||||
You can cancel a task at any time while it's running or queued. Click the **Cancel** button in the task header. A confirmation dialog appears before the task is stopped. The task transitions to `canceled` and any configured webhook fires with the canceled status.
|
||||
|
||||
<Note>
|
||||
Credits for actions already taken are still consumed. Canceling stops future actions but does not refund past ones.
|
||||
</Note>
|
||||
|
||||
---
|
||||
|
||||
## Reviewing results
|
||||
|
||||
Once a task finishes, open it from **Runs** to see the full results. The run detail page has five tabs:
|
||||
|
||||
- **Overview**: The AI's reasoning timeline alongside browser screenshots. Each Thought, Block, and Action card shows what the agent saw and why it acted.
|
||||
- **Output**: The complete JSON output and any downloaded files.
|
||||
- **Parameters**: The configuration you submitted: URL, prompt, engine, proxy location, webhook URL, data schema, and other settings.
|
||||
- **Recording**: Full video replay of the browser session. Every task is recorded automatically.
|
||||
- **Code**: Auto-generated Python code to reproduce this task via the API or SDK (when code generation is enabled).
|
||||
|
||||
For a full walkthrough of each tab, see [Run Details](/cloud/viewing-results/run-details).
|
||||
@@ -0,0 +1,94 @@
|
||||
---
|
||||
title: UI Overview
|
||||
slug: cloud/overview
|
||||
subtitle: Navigate the Skyvern Cloud dashboard
|
||||
description: Tour the Skyvern Cloud dashboard
|
||||
keywords:
|
||||
- dashboard
|
||||
- navigation
|
||||
- sidebar
|
||||
- settings
|
||||
- runs page
|
||||
- workflows page
|
||||
- run
|
||||
---
|
||||
|
||||
Skyvern Cloud ([app.skyvern.com](https://app.skyvern.com)) lets you automate any website from your browser. Describe what you want in plain English, watch an AI-powered browser do it live, and get structured results back. No code required.
|
||||
|
||||
<Tip>
|
||||
New to Skyvern? The [homepage](/) covers how Skyvern works under the hood, built-in capabilities, and the eight core concepts behind every automation.
|
||||
</Tip>
|
||||
|
||||
<Note>
|
||||
Looking to integrate Skyvern into your own app? See the [Quickstart](/developers/getting-started/quickstart) instead.
|
||||
</Note>
|
||||
|
||||
## The dashboard
|
||||
|
||||
Sign in and you'll land on the **Discover** page, the starting point for running automations.
|
||||
|
||||
<img src="/images/cloud/skyvern-cloud-discover.png" alt="Skyvern Cloud dashboard showing the Discover page" />
|
||||
|
||||
The left sidebar is your navigation hub. Here's what each section does:
|
||||
|
||||
### Build
|
||||
|
||||
Where you create and monitor automations.
|
||||
|
||||
| Page | Purpose |
|
||||
|------|---------|
|
||||
| **Discover** | Run one-off tasks. Type your instructions and target URL into a single prompt, pick an engine, and hit send. |
|
||||
| **Workflows** | Build multi-step automations with the visual workflow editor. Add loops, conditionals, and data passing between steps. |
|
||||
| **Runs** | Execution history for every task and workflow. Filter by status, drill into any run to see actions, recordings, and extracted data. |
|
||||
| **Browsers** | Active [browser sessions](/cloud/browser-management/browser-sessions). Keep a browser open across multiple tasks to preserve login state, cookies, and page context. Load a [browser profile](/cloud/browser-management/browser-profiles) to restore previously saved state. |
|
||||
|
||||
### Agents
|
||||
|
||||
Ready-made automation templates. Each agent is preconfigured with a prompt, target URL, and settings. Pick one to see it work or use it as a starting point for your own task.
|
||||
|
||||
### General
|
||||
|
||||
| Page | Purpose |
|
||||
|------|---------|
|
||||
| **Billing** | Usage, remaining credits, and plan management. |
|
||||
| **Credentials** | Store website logins, credit card details, and secret values securely. Skyvern uses these to authenticate automatically, fill payment forms, and inject secrets into workflows. |
|
||||
| **Settings** | API key, account preferences, and organization management. |
|
||||
|
||||
## How it works
|
||||
|
||||
Every automation in Skyvern Cloud follows the same pattern:
|
||||
|
||||
<Steps>
|
||||
<Step title="Describe your task">
|
||||
Type what you want into the prompt bar. Include the target URL and your instructions in one go. Something like "Get the top post from https://news.ycombinator.com" or "Fill out the contact form at https://example.com/contact with my details."
|
||||
</Step>
|
||||
<Step title="Watch it happen">
|
||||
A cloud browser opens and you see it navigate in real time. Pages load, elements highlight, actions fire. An agent log streams the AI's reasoning (what it sees on the page, what it plans to do, and why) so you can follow along. If the AI gets stuck, hit **Take Control** to jump in and help.
|
||||
</Step>
|
||||
<Step title="Get your results">
|
||||
Extracted data appears as structured JSON on the run detail page. Every run also includes an output view, full recording, the parameters you submitted, and auto-generated code to reproduce the task via API.
|
||||
</Step>
|
||||
</Steps>
|
||||
|
||||
That's it. The next guide walks you through this flow with a real example.
|
||||
|
||||
---
|
||||
|
||||
## Next steps
|
||||
|
||||
<CardGroup cols={2}>
|
||||
<Card
|
||||
title="Run Your First Task"
|
||||
icon="play"
|
||||
href="/cloud/getting-started/run-your-first-task"
|
||||
>
|
||||
Follow along with a real example to see Skyvern Cloud in action
|
||||
</Card>
|
||||
<Card
|
||||
title="Core Concepts"
|
||||
icon="book"
|
||||
href="/developers/getting-started/core-concepts"
|
||||
>
|
||||
Understand tasks, workflows, and other foundational concepts
|
||||
</Card>
|
||||
</CardGroup>
|
||||
@@ -0,0 +1,195 @@
|
||||
---
|
||||
title: The Discover Page
|
||||
subtitle: Run ad-hoc browser automations with natural language
|
||||
description: Use the Discover page in the Skyvern Cloud UI to run ad-hoc browser automation tasks with natural language prompts, choose an engine, and configure advanced settings like proxies and webhooks.
|
||||
slug: cloud/run-a-task
|
||||
keywords:
|
||||
- discover page
|
||||
- prompt
|
||||
- engine
|
||||
- proxy
|
||||
- webhook
|
||||
- data extraction
|
||||
- advanced settings
|
||||
- 2FA
|
||||
---
|
||||
|
||||
The **Discover** page is where you run one-off browser automations. Type what you want done in plain language, and Skyvern opens a browser and does it for you.
|
||||
|
||||
<img src="/images/cloud/discover-page-overview.png" alt="Discover page overview" />
|
||||
|
||||
---
|
||||
|
||||
## The prompt box
|
||||
|
||||
Type a natural language instruction describing what you want automated. Be specific about the goal and any data you want extracted.
|
||||
|
||||
**Examples:**
|
||||
- "Go to amazon.com and find the price of the MacBook Air M4"
|
||||
- "Fill out the contact form at example.com/contact with name John Doe and email john@example.com"
|
||||
- "Get an insurance quote from geico.com for a 2020 Toyota Camry"
|
||||
|
||||
<img src="/images/cloud/prompt-box-filled.png" alt="Prompt box with a sample prompt" />
|
||||
|
||||
Click the **send button** or press **Enter** to start.
|
||||
|
||||
Below the prompt box, **quick-action buttons** offer pre-built examples like "Add a product to cart" or "Get an insurance quote." Click one to run it immediately or use it as a starting point.
|
||||
|
||||
---
|
||||
|
||||
## Choosing an engine
|
||||
|
||||
The dropdown next to the send button controls which engine runs the task.
|
||||
|
||||
| Engine | Best for |
|
||||
|--------|----------|
|
||||
| **Skyvern 2.0 with Code** | Complex, multi-step tasks. Generates reusable scripts. **(Default)** |
|
||||
| **Skyvern 2.0** | Complex tasks without script generation |
|
||||
| **Skyvern 1.0** | Simple, single-objective tasks. Faster and cheaper. |
|
||||
|
||||
<Tip>
|
||||
Start with the default. Switch to Skyvern 1.0 when you have a straightforward, single-page task and want faster execution.
|
||||
</Tip>
|
||||
|
||||
---
|
||||
|
||||
## Advanced settings
|
||||
|
||||
Click the **gear icon** next to the prompt box to expand the settings panel.
|
||||
|
||||
<img src="/images/cloud/advanced-settings-panel.png" alt="Advanced settings panel" />
|
||||
|
||||
| Setting | What it does |
|
||||
|---------|-------------|
|
||||
| **Proxy Location** | Route the browser through a residential proxy in a specific country. Default is `RESIDENTIAL` (US). Set to `NONE` to disable. Available: US, UK, Germany, France, Spain, Ireland, India, Japan, Australia, Canada, Brazil, Mexico, Argentina, New Zealand, South Africa, Italy, Netherlands, Philippines, Turkey. |
|
||||
| **Webhook URL** | URL that receives a POST request when the task finishes. The payload includes status, extracted data, screenshots, and recording URL. |
|
||||
| **Browser Session ID** | Run inside an existing persistent browser session (`pbs_xxx`). Preserves cookies and login state across multiple tasks. |
|
||||
| **CDP Address** | Connect to your own browser via Chrome DevTools Protocol (e.g., `http://127.0.0.1:9222`). For local development. |
|
||||
| **2FA Identifier** | Links your TOTP credentials to this task. Skyvern uses it to retrieve the correct code when a 2FA prompt appears. |
|
||||
| **Extra HTTP Headers** | Custom headers sent with every browser request, as JSON (e.g., `{"Authorization": "Bearer token"}`). |
|
||||
| **Publish Workflow** | Save a reusable workflow alongside the task run. Re-run the same automation later from the Workflows page. |
|
||||
| **Max Steps Override** | Cap the number of AI reasoning steps. Each step = one screenshot-analyze-act cycle. Useful for controlling cost during development. |
|
||||
| **Max Screenshot Scrolls** | Number of scrolls for post-action screenshots. Increase for pages with lazy-loaded content. `0` = viewport only. |
|
||||
|
||||
---
|
||||
|
||||
## Data extraction schema
|
||||
|
||||
The **Data Schema** field in advanced settings lets you define the structure of extracted output as [JSON Schema](https://json-schema.org/).
|
||||
|
||||
Without a schema, the AI returns data in whatever format it chooses. With a schema, output conforms to your structure, making it predictable for downstream use.
|
||||
|
||||
<img src="/images/cloud/data-schema-field.png" alt="Data schema field with JSON" />
|
||||
|
||||
```json
|
||||
{
|
||||
"type": "object",
|
||||
"properties": {
|
||||
"product_name": {
|
||||
"type": "string",
|
||||
"description": "The name of the product"
|
||||
},
|
||||
"price": {
|
||||
"type": "number",
|
||||
"description": "The price in USD"
|
||||
},
|
||||
"in_stock": {
|
||||
"type": "boolean",
|
||||
"description": "Whether the product is in stock"
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
Use the `description` field on each property to guide the AI on what to extract.
|
||||
|
||||
<Accordion title="Example: Extracting a list of items">
|
||||
```json
|
||||
{
|
||||
"type": "object",
|
||||
"properties": {
|
||||
"quotes": {
|
||||
"type": "array",
|
||||
"items": {
|
||||
"type": "object",
|
||||
"properties": {
|
||||
"premium_amount": {
|
||||
"type": "string",
|
||||
"description": "Total premium in USD (e.g., '$321.57')"
|
||||
},
|
||||
"coverage_type": {
|
||||
"type": "string",
|
||||
"description": "Type of coverage (e.g., 'Full Coverage')"
|
||||
},
|
||||
"deductible": {
|
||||
"type": "string",
|
||||
"description": "Deductible amount"
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
</Accordion>
|
||||
|
||||
---
|
||||
|
||||
## Workflow templates
|
||||
|
||||
Below the prompt box, the Discover page shows a gallery of **workflow templates**: pre-built automations for common use cases.
|
||||
|
||||
<img src="/images/cloud/workflow-templates.png" alt="Workflow template gallery" />
|
||||
|
||||
Click any template to launch it with pre-filled configuration, or use it as a starting point and customize.
|
||||
|
||||
---
|
||||
|
||||
## Tips for better results
|
||||
|
||||
**Write specific prompts.** Include the exact goal, target fields, and what "done" looks like.
|
||||
|
||||
| Instead of | Write |
|
||||
|-----------|-------|
|
||||
| "Get some data from this site" | "Extract the product name, price, and availability from the first 5 results on amazon.com/s?k=wireless+mouse" |
|
||||
| "Fill out the form" | "Fill the contact form at example.com/contact with name 'Jane Doe', email 'jane@example.com', and message 'Demo request'" |
|
||||
|
||||
**Control cost with Max Steps.** Set **Max Steps Override** to a reasonable limit (e.g., 10–20 for simple tasks) during development. Each step consumes one credit. Remove the cap once you've confirmed the task works.
|
||||
|
||||
**Debug failures in order.** If a task fails or produces wrong results:
|
||||
|
||||
1. Check the **Failure Reason** at the top of the run detail page
|
||||
2. Read the **Thought cards** in the Overview timeline to find where the AI went off track
|
||||
3. Watch the **Recording** to see what actually happened on screen
|
||||
4. Review **Parameters** to confirm the inputs were correct
|
||||
|
||||
---
|
||||
|
||||
## What happens next
|
||||
|
||||
1. Your prompt is sent to Skyvern
|
||||
2. A cloud browser opens and navigates to the target URL (or finds one from your prompt)
|
||||
3. The AI analyzes the page, plans actions, and executes them step by step
|
||||
4. You're taken to the [live execution view](/cloud/getting-started/monitor-a-run) where you can watch it happen in real time
|
||||
5. When complete, results appear on the run detail page under **Runs**
|
||||
|
||||
---
|
||||
|
||||
## Next steps
|
||||
|
||||
<CardGroup cols={2}>
|
||||
<Card
|
||||
title="Watching Live Execution"
|
||||
icon="eye"
|
||||
href="/cloud/getting-started/monitor-a-run"
|
||||
>
|
||||
Monitor runs, take control of the browser, and review results
|
||||
</Card>
|
||||
<Card
|
||||
title="Build a Workflow"
|
||||
icon="diagram-project"
|
||||
href="/cloud/building-workflows/build-a-workflow"
|
||||
>
|
||||
Turn a successful task into a reusable multi-step workflow
|
||||
</Card>
|
||||
</CardGroup>
|
||||
@@ -0,0 +1,357 @@
|
||||
---
|
||||
title: Run from Code
|
||||
subtitle: Execute tasks and retrieve results via the API or SDK
|
||||
description: Run browser automation tasks programmatically using the Python SDK, TypeScript SDK, or REST API. Poll for results, use webhooks, or wait for completion.
|
||||
slug: cloud/getting-started/run-from-code
|
||||
keywords:
|
||||
- run_task
|
||||
- get_run
|
||||
- polling
|
||||
- wait_for_completion
|
||||
- webhook
|
||||
- SDK
|
||||
- API
|
||||
---
|
||||
|
||||
Run tasks programmatically using the Python SDK, TypeScript SDK, or REST API.
|
||||
|
||||
---
|
||||
|
||||
## Run a task
|
||||
|
||||
A task has one required parameter and one commonly used optional parameter:
|
||||
|
||||
- **`prompt`** (required): Natural language instructions describing what the AI should do
|
||||
- **`url`** (optional): The starting page for the automation
|
||||
|
||||
When you call `run_task`, Skyvern spins up a cloud browser, navigates to the URL, and executes your prompt. A typical task takes 30-90 seconds depending on complexity.
|
||||
|
||||
<CodeGroup>
|
||||
```python Python
|
||||
import os
|
||||
import asyncio
|
||||
from skyvern import Skyvern
|
||||
|
||||
async def main():
|
||||
client = Skyvern(api_key=os.getenv("SKYVERN_API_KEY"))
|
||||
|
||||
result = await client.run_task(
|
||||
prompt="Get the title of the top post",
|
||||
url="https://news.ycombinator.com",
|
||||
)
|
||||
|
||||
print(f"Run ID: {result.run_id}")
|
||||
print(f"Status: {result.status}")
|
||||
|
||||
asyncio.run(main())
|
||||
```
|
||||
|
||||
```typescript TypeScript
|
||||
import { SkyvernClient } from "@skyvern/client";
|
||||
|
||||
async function main() {
|
||||
const client = new SkyvernClient({
|
||||
apiKey: process.env.SKYVERN_API_KEY,
|
||||
});
|
||||
|
||||
const result = await client.runTask({
|
||||
body: {
|
||||
prompt: "Get the title of the top post",
|
||||
url: "https://news.ycombinator.com",
|
||||
},
|
||||
});
|
||||
|
||||
console.log(`Run ID: ${result.run_id}`);
|
||||
console.log(`Status: ${result.status}`);
|
||||
}
|
||||
|
||||
main();
|
||||
```
|
||||
|
||||
```bash cURL
|
||||
curl -X POST "https://api.skyvern.com/v1/run/tasks" \
|
||||
-H "x-api-key: $SKYVERN_API_KEY" \
|
||||
-H "Content-Type: application/json" \
|
||||
-d '{
|
||||
"prompt": "Get the title of the top post",
|
||||
"url": "https://news.ycombinator.com"
|
||||
}'
|
||||
```
|
||||
</CodeGroup>
|
||||
|
||||
**Example response:**
|
||||
|
||||
```json
|
||||
{
|
||||
"run_id": "tsk_v2_486305187432193504",
|
||||
"status": "queued",
|
||||
"output": null,
|
||||
"downloaded_files": null,
|
||||
"recording_url": null,
|
||||
"screenshot_urls": null,
|
||||
"failure_reason": null,
|
||||
"created_at": "2026-01-20T11:52:29.276851",
|
||||
"modified_at": "2026-01-20T11:52:29.484284",
|
||||
"app_url": "https://app.skyvern.com/runs/wr_486305187432193510",
|
||||
"run_type": "task_v2"
|
||||
}
|
||||
```
|
||||
|
||||
The response includes a `run_id`. Use this ID to check status, fetch results, and retrieve artifacts.
|
||||
|
||||
<Warning>
|
||||
`run_task` returns immediately -- the task is queued, not finished. Always poll or use webhooks to get results.
|
||||
</Warning>
|
||||
|
||||
---
|
||||
|
||||
## Get results
|
||||
|
||||
The `run_task` call queues the task and returns immediately.
|
||||
|
||||
Use the `run_id` to fetch results once the task reaches a terminal state.
|
||||
|
||||
| Status | Description |
|
||||
|--------|-------------|
|
||||
| `created` | Task initialized, not yet queued |
|
||||
| `queued` | Waiting for an available browser |
|
||||
| `running` | AI is navigating and executing |
|
||||
| `completed` | Task finished successfully -- check `output` for results |
|
||||
| `failed` | Task encountered an error -- check `failure_reason` and retry or adjust your prompt |
|
||||
| `terminated` | Task was manually stopped |
|
||||
| `timed_out` | Task exceeded time limit -- increase `max_steps` or simplify the task |
|
||||
| `canceled` | Task was canceled before starting |
|
||||
|
||||
You have three options for retrieving results:
|
||||
|
||||
### Option 1: Polling
|
||||
|
||||
Poll `get_run` until status is terminal (`completed`, `failed`, `terminated`, `timed_out`, or `canceled`).
|
||||
|
||||
<CodeGroup>
|
||||
```python Python
|
||||
run_id = result.run_id
|
||||
|
||||
while True:
|
||||
run = await client.get_run(run_id)
|
||||
|
||||
if run.status in ["completed", "failed", "terminated", "timed_out", "canceled"]:
|
||||
break
|
||||
|
||||
await asyncio.sleep(5)
|
||||
|
||||
print(f"Output: {run.output}")
|
||||
```
|
||||
|
||||
```typescript TypeScript
|
||||
const runId = result.run_id;
|
||||
|
||||
while (true) {
|
||||
const run = await client.getRun(runId);
|
||||
|
||||
if (["completed", "failed", "terminated", "timed_out", "canceled"].includes(run.status)) {
|
||||
console.log(`Output: ${JSON.stringify(run.output)}`);
|
||||
break;
|
||||
}
|
||||
|
||||
await new Promise((resolve) => setTimeout(resolve, 5000));
|
||||
}
|
||||
```
|
||||
|
||||
```bash cURL
|
||||
#!/bin/bash
|
||||
RUN_ID="YOUR_RUN_ID"
|
||||
|
||||
while true; do
|
||||
RESPONSE=$(curl -s -X GET "https://api.skyvern.com/v1/runs/$RUN_ID" \
|
||||
-H "x-api-key: $SKYVERN_API_KEY")
|
||||
|
||||
STATUS=$(echo "$RESPONSE" | jq -r '.status')
|
||||
echo "Status: $STATUS"
|
||||
|
||||
if [[ "$STATUS" == "completed" || "$STATUS" == "failed" || "$STATUS" == "terminated" || "$STATUS" == "timed_out" || "$STATUS" == "canceled" ]]; then
|
||||
echo "$RESPONSE" | jq '.output'
|
||||
break
|
||||
fi
|
||||
|
||||
sleep 5
|
||||
done
|
||||
```
|
||||
</CodeGroup>
|
||||
|
||||
<Warning>
|
||||
Your polling loop must check **all** terminal states: `completed`, `failed`, `terminated`, `timed_out`, `canceled`. Missing one causes infinite loops.
|
||||
</Warning>
|
||||
|
||||
### Option 2: Webhooks
|
||||
|
||||
Pass a `webhook_url` when creating the task. Skyvern sends a POST request to your URL when the task completes.
|
||||
|
||||
<CodeGroup>
|
||||
```python Python
|
||||
result = await client.run_task(
|
||||
prompt="Get the title of the top post",
|
||||
url="https://news.ycombinator.com",
|
||||
webhook_url="https://your-server.com/webhook",
|
||||
)
|
||||
```
|
||||
|
||||
```typescript TypeScript
|
||||
const result = await client.runTask({
|
||||
body: {
|
||||
prompt: "Get the title of the top post",
|
||||
url: "https://news.ycombinator.com",
|
||||
webhook_url: "https://your-server.com/webhook",
|
||||
},
|
||||
});
|
||||
```
|
||||
|
||||
```bash cURL
|
||||
curl -X POST "https://api.skyvern.com/v1/run/tasks" \
|
||||
-H "x-api-key: $SKYVERN_API_KEY" \
|
||||
-H "Content-Type: application/json" \
|
||||
-d '{
|
||||
"prompt": "Get the title of the top post",
|
||||
"url": "https://news.ycombinator.com",
|
||||
"webhook_url": "https://your-server.com/webhook"
|
||||
}'
|
||||
```
|
||||
</CodeGroup>
|
||||
|
||||
Skyvern sends a POST request with the full run data when the task completes or fails.
|
||||
|
||||
### Option 3: Wait for completion (Python only)
|
||||
|
||||
Block until the task finishes instead of polling manually.
|
||||
|
||||
```python Python
|
||||
result = await client.run_task(
|
||||
prompt="Get the title of the top post",
|
||||
url="https://news.ycombinator.com",
|
||||
wait_for_completion=True,
|
||||
)
|
||||
print(result.output)
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Understand the response
|
||||
|
||||
The response from polling (`get_run`) and webhooks have slightly different structures. Both contain the core task data, but webhooks include additional metadata.
|
||||
|
||||
<CodeGroup>
|
||||
```json Polling Response
|
||||
{
|
||||
"run_id": "tsk_v2_486305187432193504",
|
||||
"status": "completed",
|
||||
"output": {
|
||||
"top_post_title": "Linux kernel framework for PCIe device emulation, in userspace"
|
||||
},
|
||||
"downloaded_files": [],
|
||||
"recording_url": "https://skyvern-artifacts.s3.amazonaws.com/v1/production/.../recording.webm?...",
|
||||
"screenshot_urls": ["https://skyvern-artifacts.s3.amazonaws.com/.../screenshot_final.png?..."],
|
||||
"failure_reason": null,
|
||||
"errors": [],
|
||||
"step_count": 2,
|
||||
"run_type": "task_v2",
|
||||
"app_url": "https://app.skyvern.com/runs/wr_486305187432193510",
|
||||
"browser_session_id": null,
|
||||
"browser_profile_id": null,
|
||||
"created_at": "2026-01-20T11:52:29.276851",
|
||||
"modified_at": "2026-01-20T11:54:08.822807",
|
||||
"queued_at": "2026-01-20T11:52:29.483922",
|
||||
"started_at": "2026-01-20T11:52:31.835337",
|
||||
"finished_at": "2026-01-20T11:54:08.821985",
|
||||
"run_request": {
|
||||
"prompt": "Get the title of the top post",
|
||||
"url": "https://news.ycombinator.com/",
|
||||
"engine": "skyvern-2.0"
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
```json Webhook Payload
|
||||
{
|
||||
"task_id": "tsk_v2_486306851394503256",
|
||||
"run_id": "tsk_v2_486306851394503256",
|
||||
"status": "completed",
|
||||
"output": {
|
||||
"top_post_title": "Linux kernel framework for PCIe device emulation, in userspace"
|
||||
},
|
||||
"summary": "I have successfully navigated to Hacker News and extracted the title of the top post.",
|
||||
"prompt": "Get the title of the top post",
|
||||
"url": "https://news.ycombinator.com/",
|
||||
"downloaded_files": [],
|
||||
"recording_url": "https://skyvern-artifacts.s3.amazonaws.com/v1/production/.../recording.webm?...",
|
||||
"screenshot_urls": ["https://skyvern-artifacts.s3.amazonaws.com/.../screenshot_final.png?..."],
|
||||
"failure_reason": null,
|
||||
"errors": [],
|
||||
"step_count": 2,
|
||||
"run_type": "task_v2",
|
||||
"app_url": "https://app.skyvern.com/runs/wr_486306851394503262",
|
||||
"organization_id": "o_485917350850524254",
|
||||
"workflow_run_id": "wr_486306851394503262",
|
||||
"proxy_location": "RESIDENTIAL",
|
||||
"webhook_callback_url": "https://your-server.com/webhook",
|
||||
"created_at": "2026-01-20T11:58:57.414123",
|
||||
"modified_at": "2026-01-20T12:00:31.513449",
|
||||
"queued_at": "2026-01-20T11:58:57.607233",
|
||||
"started_at": "2026-01-20T11:58:59.395028",
|
||||
"finished_at": "2026-01-20T12:00:31.512692"
|
||||
}
|
||||
```
|
||||
</CodeGroup>
|
||||
|
||||
**Common fields (both polling and webhook):**
|
||||
|
||||
| Field | Type | Description |
|
||||
|-------|------|-------------|
|
||||
| `run_id` | string | Unique identifier for this run |
|
||||
| `status` | string | Current status: `queued`, `running`, `completed`, `failed`, `terminated`, `timed_out`, `canceled` |
|
||||
| `output` | object \| null | Extracted data from the task |
|
||||
| `downloaded_files` | array | Files downloaded during execution |
|
||||
| `recording_url` | string \| null | Video recording of the browser session |
|
||||
| `screenshot_urls` | array \| null | Screenshots captured (latest first) |
|
||||
| `failure_reason` | string \| null | Error message if the run failed |
|
||||
| `errors` | array | List of errors encountered |
|
||||
| `step_count` | integer \| null | Number of steps executed |
|
||||
| `run_type` | string | Type of run: `task_v2`, `openai_cua`, `anthropic_cua` |
|
||||
| `app_url` | string | Link to view this run in Skyvern Cloud |
|
||||
| `created_at` | datetime | When the run was created |
|
||||
| `modified_at` | datetime | When the run was last updated |
|
||||
| `queued_at` | datetime \| null | When the run entered the queue |
|
||||
| `started_at` | datetime \| null | When execution began |
|
||||
| `finished_at` | datetime \| null | When execution completed |
|
||||
|
||||
**Polling-only fields:**
|
||||
|
||||
| Field | Type | Description |
|
||||
|-------|------|-------------|
|
||||
| `run_request` | object | Original request parameters (prompt, url, engine, etc.) |
|
||||
| `browser_session_id` | string \| null | ID of the browser session used |
|
||||
| `browser_profile_id` | string \| null | ID of the browser profile used |
|
||||
| `max_screenshot_scrolls` | integer \| null | Number of scrolls for screenshots |
|
||||
| `script_run` | object \| null | Script run result if AI fallback triggered |
|
||||
|
||||
**Webhook-only fields:**
|
||||
|
||||
| Field | Type | Description |
|
||||
|-------|------|-------------|
|
||||
| `task_id` | string | Same as `run_id` |
|
||||
| `summary` | string | AI-generated description of what was done |
|
||||
| `prompt` | string | The prompt from the original request |
|
||||
| `url` | string | The URL from the original request |
|
||||
| `organization_id` | string | Your organization ID |
|
||||
| `workflow_run_id` | string | Associated workflow run ID |
|
||||
| `proxy_location` | string | Proxy location used (e.g., `RESIDENTIAL`) |
|
||||
| `webhook_callback_url` | string | The webhook URL that received this payload |
|
||||
|
||||
### Artifacts
|
||||
|
||||
Every run captures recordings, screenshots, and logs. See [Using Artifacts](/developers/debugging/using-artifacts) for retrieval and the full artifact type reference.
|
||||
|
||||
---
|
||||
|
||||
<Note>
|
||||
For multi-step automations that chain multiple actions, see [Browser Automation](/developers/browser-automations/overview).
|
||||
</Note>
|
||||
@@ -0,0 +1,141 @@
|
||||
---
|
||||
title: Your First Task
|
||||
slug: cloud/run-your-first-task
|
||||
subtitle: Run a browser automation from start to finish
|
||||
description: Follow a step-by-step tutorial to run your first browser automation task in the Skyvern Cloud UI, from writing a prompt to viewing extracted results.
|
||||
keywords:
|
||||
- first task
|
||||
- tutorial
|
||||
- discover page
|
||||
- prompt
|
||||
- extracted data
|
||||
- live viewer
|
||||
---
|
||||
|
||||
Let's run a real automation. You'll tell Skyvern to visit a website, extract data, and return it as JSON. Then watch the entire thing happen live.
|
||||
|
||||
## Step 1: Write your prompt
|
||||
|
||||
Open [app.skyvern.com](https://app.skyvern.com) and you'll land on the **Discover** page.
|
||||
|
||||
<img src="/images/cloud/skyvern-cloud-discover.png" alt="Discover page with a prompt entered" />
|
||||
|
||||
The Discover page has a single input field. Type your instructions and include the target URL in the same prompt. For this example, enter:
|
||||
|
||||
```
|
||||
Get the title of the #1 post on the front page for https://news.ycombinator.com
|
||||
```
|
||||
|
||||
That's it. Skyvern parses the URL and figures out how to navigate the page and extract the data.
|
||||
|
||||
Below the input, you'll see quick-action chips like "Add a product to cart" and "What's the top post on hackernews". Click any of these to try a pre-filled example instead.
|
||||
|
||||
<Tip>
|
||||
The more specific your prompt, the better. "Get the title of the #1 post" works much better than "get some data." Include the exact fields you want, what success looks like, and any constraints.
|
||||
</Tip>
|
||||
|
||||
## Step 2: Pick an engine and run
|
||||
|
||||
Next to your prompt, you'll see an engine selector. Click it to switch engines:
|
||||
|
||||
| Engine | When to use it |
|
||||
|--------|---------------|
|
||||
| **Skyvern 1.0** | Tasks with a simple, single goal: filling a form, searching for information on Google, reading content from a page |
|
||||
| **Skyvern 2.0** | Complex, multi-step tasks. Scores state-of-the-art 85.85% on the WebVoyager benchmark |
|
||||
| **Skyvern 2.0 with Code** | The default engine. Same capabilities as Skyvern 2.0, plus auto-generates reusable code and a workflow from the run |
|
||||
|
||||
For this example, keep the default **Skyvern 2.0 with Code** selected.
|
||||
|
||||
Click the **send button** (arrow icon to the right of the input). Skyvern generates a workflow from your prompt and opens it in the workflow editor. Click **Run** in the top right, confirm the parameters, then click **Run workflow** to start execution.
|
||||
|
||||
<Accordion title="Optional: Advanced settings">
|
||||
Click the **gear icon** next to send to configure additional options before running:
|
||||
|
||||
| Setting | What it does |
|
||||
|---------|-------------|
|
||||
| **Webhook Callback URL** | Endpoint to receive the extracted data when the run completes |
|
||||
| **Proxy Location** | Route Skyvern through one of the available proxies |
|
||||
| **Browser Session ID** | Reuse a persistent browser session to keep login state |
|
||||
| **CDP Address** | Connect to your own browser via Chrome DevTools Protocol |
|
||||
| **2FA Identifier** | Identifier for a 2FA code to handle two-factor auth automatically |
|
||||
| **Extra HTTP Headers** | Custom HTTP request headers (dict format) |
|
||||
| **Generate Script** | Auto-generate reusable scripts from a successful run |
|
||||
| **Publish Workflow** | Create a workflow alongside this task run |
|
||||
| **Max Steps Override** | Cap the number of steps the AI can take |
|
||||
| **Data Schema** | Define structured JSON output format |
|
||||
| **Max Screenshot Scrolls** | Limit scrolls for post-action screenshots (default: 3) |
|
||||
|
||||
These are all optional. The defaults work for most tasks.
|
||||
</Accordion>
|
||||
|
||||
## Step 3: Watch the live browser
|
||||
|
||||
This is where it gets interesting. Once the task starts, you'll see the run detail page with a live view of the browser:
|
||||
|
||||
<img src="/images/cloud/discover-prompt-in-process.png" alt="Run detail page showing a live browser navigating Hacker News" />
|
||||
|
||||
On the left, a **live browser view**. You'll see pages load, elements highlight, and actions fire.
|
||||
|
||||
On the right, the **agent log**. A running stream of the AI's Thoughts, Decisions, and block executions. If something goes wrong, this is where you'll figure out why.
|
||||
|
||||
## Step 4: Review the results
|
||||
|
||||
When the task finishes, the status badge flips to **completed** and the extracted data appears at the top of the page.
|
||||
|
||||
<img src="/images/cloud/discover-workflow-completed.png" alt="Completed run showing extracted data and result tabs" />
|
||||
|
||||
### Extracted data
|
||||
|
||||
The **Extracted Information** block shows your results as structured JSON:
|
||||
|
||||
```json
|
||||
[
|
||||
{
|
||||
"top_post_title": "Don't rent the cloud, own instead"
|
||||
}
|
||||
]
|
||||
```
|
||||
|
||||
Your result will differ, since the #1 post changes constantly. The structure is what matters.
|
||||
|
||||
The agent log on the right confirms what happened. You'll see a final Thought summarizing the result.
|
||||
|
||||
### Tabs
|
||||
|
||||
Below the extracted data, five tabs give you different views of the run:
|
||||
|
||||
- **Overview**: The AI's reasoning timeline alongside browser screenshots. Each Thought, Block, and Action card shows what the agent saw and why it acted.
|
||||
- **Output**: The complete JSON output and any downloaded files.
|
||||
- **Parameters**: The exact configuration that was submitted (URL, prompt, engine, schema). Useful for reproducing or tweaking the run.
|
||||
- **Recording**: Full video replay of the browser session, start to finish.
|
||||
- **Code**: Auto-generated Python code to reproduce this task via the API or SDK.
|
||||
|
||||
## Try something bigger
|
||||
|
||||
Now that you've seen the basic flow, here are a few ideas to try next:
|
||||
|
||||
- **Fill a form**: Point Skyvern at a contact form and tell it what to enter in each field
|
||||
- **Compare prices**: Extract product names and prices from an e-commerce page using a data schema
|
||||
- **Navigate a flow**: Use Skyvern 2.0 to walk through a multi-page checkout or signup process
|
||||
- **Use an Agent template**: Check the **Agents** section in the sidebar for pre-built automations you can run instantly
|
||||
|
||||
---
|
||||
|
||||
## Next steps
|
||||
|
||||
<CardGroup cols={2}>
|
||||
<Card
|
||||
title="Run a Task via API"
|
||||
icon="code"
|
||||
href="/running-automations/run-a-task"
|
||||
>
|
||||
Trigger automations programmatically with the Skyvern API
|
||||
</Card>
|
||||
<Card
|
||||
title="Core Concepts"
|
||||
icon="book"
|
||||
href="/developers/getting-started/core-concepts"
|
||||
>
|
||||
Understand tasks, workflows, and other building blocks
|
||||
</Card>
|
||||
</CardGroup>
|
||||
Reference in New Issue
Block a user