Dorod-Sky/docs/sdk-reference/overview.mdx

---
title: Python SDK
subtitle: Install, authenticate, and configure the Skyvern Python client
slug: sdk-reference/overview
---

The `skyvern` package wraps the Skyvern REST API in a typed, async Python client.

```bash
pip install skyvern
```

<Note>
Requires Python 3.11+. If you hit version errors, use `pipx install skyvern` to install in an isolated environment.
</Note>

---

## Initialize the client

The `Skyvern` class is async — all methods are coroutines. Wrap calls in an `async` function and use `asyncio.run()` as the entry point:

```python
import asyncio
from skyvern import Skyvern

async def main():
    client = Skyvern(api_key="YOUR_API_KEY")

    result = await client.run_task(
        prompt="Get the title of the top post on Hacker News",
        url="https://news.ycombinator.com",
        wait_for_completion=True,
    )
    print(result.output)

asyncio.run(main())
```

If you're inside a framework that already runs an event loop (FastAPI, Django ASGI), `await` directly without `asyncio.run()`.

### Constructor parameters

| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `api_key` | `str` | — | **Required.** Your Skyvern API key. Get one at [app.skyvern.com/settings](https://app.skyvern.com/settings/api-keys). |
| `environment` | `SkyvernEnvironment` | `CLOUD` | Target environment. Options: `CLOUD`, `STAGING`, `LOCAL`. |
| `base_url` | `str \| None` | `None` | Override the API base URL. Use this for self-hosted deployments. |
| `timeout` | `float \| None` | `None` | HTTP request timeout in seconds. |
| `follow_redirects` | `bool` | `True` | Whether to follow HTTP redirects. |
| `httpx_client` | `AsyncClient \| None` | `None` | Provide your own httpx client for custom TLS, proxying, or connection pooling. |

---

## Environments

The SDK ships with two built-in environment URLs:

```python
from skyvern.client import SkyvernEnvironment
```

| Environment | URL | When to use |
|-------------|-----|-------------|
| `SkyvernEnvironment.CLOUD` | `https://api.skyvern.com` | Skyvern Cloud (default) |
| `SkyvernEnvironment.LOCAL` | `http://localhost:8000` | Local server started with `skyvern run server` |

For a self-hosted instance at a custom URL, pass `base_url` instead:

```python
client = Skyvern(
    api_key="YOUR_API_KEY",
    base_url="https://skyvern.your-company.com",
)
```

---

## Local mode

Run Skyvern entirely on your machine — no cloud, no network calls. `Skyvern.local()` reads your `.env` file, boots the engine in-process, and connects the client to it.

**Prerequisite:** Run `skyvern quickstart` once to create the `.env` file with your database connection and LLM API keys.

```python
from skyvern import Skyvern

client = Skyvern.local()

result = await client.run_task(
    prompt="Get the title of the top post",
    url="https://news.ycombinator.com",
    wait_for_completion=True,
)
```

If you configured headful mode during `skyvern quickstart`, a Chromium window opens on your machine so you can watch the AI work.

| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `llm_config` | `LLMConfig \| LLMRouterConfig \| None` | `None` | Override the LLM. If omitted, uses `LLM_KEY` from `.env`. |
| `settings` | `dict \| None` | `None` | Override `.env` settings at runtime. Example: `{"MAX_STEPS_PER_RUN": 100}` |

---

## `wait_for_completion`

By default, `run_task` and `run_workflow` return immediately after the run is queued — you get a `run_id` and need to poll `get_run()` yourself. Pass `wait_for_completion=True` to have the SDK poll automatically until the run reaches a terminal state (`completed`, `failed`, `terminated`, `timed_out`, or `canceled`):

```python
# Returns only after the task finishes (up to 30 min by default)
result = await client.run_task(
    prompt="Fill out the contact form",
    url="https://example.com/contact",
    wait_for_completion=True,
    timeout=600,  # give up after 10 minutes
)

# Without wait_for_completion — returns immediately
task = await client.run_task(
    prompt="Fill out the contact form",
    url="https://example.com/contact",
)
print(task.run_id)  # poll with client.get_run(task.run_id)
```

| Parameter | Type | Default | Description |
|-----------|------|---------|-------------|
| `wait_for_completion` | `bool` | `False` | Poll until the run finishes. |
| `timeout` | `float` | `1800` | Maximum wait time in seconds. Raises Python's `TimeoutError` if exceeded. |

Supported on `run_task`, `run_workflow`, and `login`.

---

## Request options

Every method accepts an optional `request_options` parameter for per-request overrides of timeout, retries, and headers:

```python
from skyvern.client.core import RequestOptions

result = await client.run_task(
    prompt="Extract data",
    url="https://example.com",
    request_options=RequestOptions(
        timeout_in_seconds=120,
        max_retries=3,
        additional_headers={"x-custom-header": "value"},
    ),
)
```

These override the client-level defaults for that single call only.

---

## Next steps

<CardGroup cols={2}>
  <Card
    title="Tasks"
    icon="play"
    href="/sdk-reference/tasks"
  >
    Run browser automations with `run_task`
  </Card>
  <Card
    title="Workflows"
    icon="diagram-project"
    href="/sdk-reference/workflows"
  >
    Create and run multi-step automations
  </Card>
  <Card
    title="Browser Sessions"
    icon="browser"
    href="/sdk-reference/browser-sessions"
  >
    Maintain live browser state between calls
  </Card>
  <Card
    title="Error Handling"
    icon="triangle-exclamation"
    href="/sdk-reference/error-handling"
  >
    Handle errors and configure retries
  </Card>
</CardGroup>