Use browser actions

Browser actions change the rendered page before Context.dev captures its content. Use them for consent dialogs, “load more” buttons, tabs, lazy-loaded feeds, and scrollable panels.

Actions require a paid plan. A Markdown, HTML, or image request with actions costs 2 credits. An enriched image request costs 5 credits. Actions bypass the scrape cache.

Prerequisites

Export an API key from the dashboard:

export CONTEXT_DEV_API_KEY="ctxt_secret_..."

Scroll a lazy-loaded page

Pass up to five actions in execution order. This request scrolls at most ten viewports and stops early when the page no longer grows.

curl --get https://api.context.dev/v1/web/scrape/markdown \
  --header "Authorization: Bearer $CONTEXT_DEV_API_KEY" \
  --data-urlencode "url=https://example.com/feed" \
  --data-urlencode 'actions=[{"do":"scroll","direction":"down","amount":"viewport","maxScrolls":10}]'

import ContextDev from "context.dev";

const client = new ContextDev({apiKey: process.env.CONTEXT_DEV_API_KEY});
const url = "https://example.com/feed";

const page = await client.web.webScrapeMd({url}, {
  query: {
    url,
    actions: JSON.stringify([
      {
        "do": "scroll",
        "direction": "down",
        "amount": "viewport",
        "maxScrolls": 10
      }
    ]),
  },
});

console.log(page.actionsApplied);
console.log(page.markdown);

import json
import os
from context.dev import ContextDev

client = ContextDev(api_key=os.environ["CONTEXT_DEV_API_KEY"])

page = client.web.web_scrape_md(
    url="https://example.com/feed",
    extra_query={
        "actions": json.dumps([
          {
            "do": "scroll",
            "direction": "down",
            "amount": "viewport",
            "maxScrolls": 10
          }
        ]),
    },
)

print(page.actions_applied)
print(page.markdown)

require "cgi/core"
require "json"
require "context_dev"

client = ContextDev::Client.new(api_key: ENV.fetch("CONTEXT_DEV_API_KEY"))
actions = [{do: "scroll", direction: "down", amount: "viewport", maxScrolls: 10}]

page = client.web.web_scrape_md(
  url: "https://example.com/feed",
  request_options: {extra_query: {"actions" => [JSON.generate(actions)]}},
)

puts page.actions_applied
puts page.markdown

package main

import (
    "context"
    "fmt"
    "os"

    contextdev "github.com/context-dot-dev/context-go-sdk/v2"
    "github.com/context-dot-dev/context-go-sdk/v2/option"
)

func main() {
    client := contextdev.NewClient(
        option.WithAPIKey(os.Getenv("CONTEXT_DEV_API_KEY")),
    )

    page, err := client.Web.WebScrapeMd(
        context.Background(),
        contextdev.WebWebScrapeMdParams{URL: "https://example.com/feed"},
        option.WithQuery("actions", `[{"do":"scroll","direction":"down","amount":"viewport","maxScrolls":10}]`),
    )
    if err != nil {
        panic(err)
    }

    fmt.Println(page.ActionsApplied)
    fmt.Println(page.Markdown)
}

<?php

require __DIR__.'/vendor/autoload.php';

use ContextDev\Client;

$client = new Client(apiKey: getenv('CONTEXT_DEV_API_KEY'));

$page = $client->web->webScrapeMd(
    url: 'https://example.com/feed',
    actions: [
        ['do' => 'scroll', 'direction' => 'down', 'amount' => 'viewport', 'maxScrolls' => 10],
    ],
);

print_r($page->actionsApplied);
echo $page->markdown, PHP_EOL;

TypeScript, Python, Ruby, and Go SDK 2.14.0 need an explicit JSON query value for actions. The examples use request options to preserve the action array.

Choose an action

Action	Required fields	Optional controls	Use it for
`wait`	`do`, `timeMs`	none	A known delay after another action. Maximum 30 seconds.
`perform`	`do`, `action`	none	One natural-language click or interaction, up to 500 characters.
`scroll`	`do`	`direction`, `amount`, `container`, `maxScrolls`	Page or panel scrolling.

A bare { "do": "scroll" } scrolls one viewport down. amount can be a pixel count, viewport, or max; maxScrolls accepts 1 to 50.

Combine click and wait

Use a specific instruction that can be verified from visible page state, then wait only as long as the resulting update needs.

curl --get https://api.context.dev/v1/web/scrape/markdown \
  --header "Authorization: Bearer $CONTEXT_DEV_API_KEY" \
  --data-urlencode "url=https://example.com/results" \
  --data-urlencode 'actions=[{"do":"perform","action":"click the \"Load more\" button"},{"do":"wait","timeMs":1500}]'

import ContextDev from "context.dev";

const client = new ContextDev({apiKey: process.env.CONTEXT_DEV_API_KEY});
const url = "https://example.com/results";

const page = await client.web.webScrapeMd({url}, {
  query: {
    url,
    actions: JSON.stringify([
      {
        "do": "perform",
        "action": "click the \"Load more\" button"
      },
      {
        "do": "wait",
        "timeMs": 1500
      }
    ]),
  },
});

console.log(page.markdown);

import json
import os
from context.dev import ContextDev

client = ContextDev(api_key=os.environ["CONTEXT_DEV_API_KEY"])

page = client.web.web_scrape_md(
    url="https://example.com/results",
    extra_query={
        "actions": json.dumps([
          {
            "do": "perform",
            "action": "click the \"Load more\" button"
          },
          {
            "do": "wait",
            "timeMs": 1500
          }
        ]),
    },
)

print(page.markdown)

require "cgi/core"
require "json"
require "context_dev"

client = ContextDev::Client.new(api_key: ENV.fetch("CONTEXT_DEV_API_KEY"))
actions = [
  {do: "perform", action: 'click the "Load more" button'},
  {do: "wait", timeMs: 1500},
]

page = client.web.web_scrape_md(
  url: "https://example.com/results",
  request_options: {extra_query: {"actions" => [JSON.generate(actions)]}},
)

puts page.markdown

package main

import (
    "context"
    "fmt"
    "os"

    contextdev "github.com/context-dot-dev/context-go-sdk/v2"
    "github.com/context-dot-dev/context-go-sdk/v2/option"
)

func main() {
    client := contextdev.NewClient(
        option.WithAPIKey(os.Getenv("CONTEXT_DEV_API_KEY")),
    )

    page, err := client.Web.WebScrapeMd(
        context.Background(),
        contextdev.WebWebScrapeMdParams{URL: "https://example.com/results"},
        option.WithQuery("actions", `[{"do":"perform","action":"click the \"Load more\" button"},{"do":"wait","timeMs":1500}]`),
    )
    if err != nil {
        panic(err)
    }

    fmt.Println(page.Markdown)
}

<?php

require __DIR__.'/vendor/autoload.php';

use ContextDev\Client;

$client = new Client(apiKey: getenv('CONTEXT_DEV_API_KEY'));

$page = $client->web->webScrapeMd(
    url: 'https://example.com/results',
    actions: [
        ['do' => 'perform', 'action' => 'click the "Load more" button'],
        ['do' => 'wait', 'timeMs' => 1500],
    ],
);

echo $page->markdown, PHP_EOL;

For an inner panel, add a CSS selector:

{
  "do": "scroll",
  "container": "[data-results-panel]",
  "amount": "viewport",
  "maxScrolls": 5
}

Verify the outcome

Do not infer success only from the HTTP status. When action outcomes are returned, inspect actionsApplied in request order:

{
  "actionsApplied": [
    {
      "instruction": "click the Load more button",
      "status": "applied",
      "method": "click",
      "targetDescription": "Load more button",
      "completionEvidence": "The results list increased from 20 to 40 items",
      "durationMs": 842
    }
  ]
}

Status	Meaning
`applied`	The requested page state was visibly verified.
`failed`	The requested state was not verified; inspect `error`.
`skipped`	The action was not attempted.

If a critical action is not applied, do not trust the captured content to represent the intended state.

Keep interactions robust

Describe visible text and intent, such as click the "Pricing" tab, instead of coordinates.
Prefer one perform action followed by a short wait over several vague instructions.
Use scroll for repeated movement; do not expand a simple scroll into multiple perform actions.
Keep a fallback for login walls, CAPTCHA challenges, and interactions that depend on a user account.

Supported endpoints

Actions are available on:

Scrape Markdown

Apply actions before reading a page as Markdown.

Scrape HTML

Apply actions before capturing rendered HTML.

Scrape images

Apply actions before collecting page images.

Extract structured data

Apply actions before extracting fields with a JSON Schema.

On Extract, actions run on the starting page before link discovery. Additional crawled pages do not repeat them. See the extraction guide for its pricing and crawl controls.

Get started

Web data

Brand data

Automation

Production

Prerequisites

Scroll a lazy-loaded page

Choose an action

Combine click and wait

Verify the outcome

Keep interactions robust

Supported endpoints

Scrape Markdown

Scrape HTML

Scrape images

Extract structured data

Next steps

Scrape a webpage

Extract page images

​Prerequisites

​Scroll a lazy-loaded page

​Choose an action

​Combine click and wait

​Verify the outcome

​Keep interactions robust

​Supported endpoints

Scrape Markdown

Scrape HTML

Scrape images

Extract structured data

​Next steps

Scrape a webpage

Extract page images

Prerequisites

Scroll a lazy-loaded page

Choose an action

Combine click and wait

Verify the outcome

Keep interactions robust

Supported endpoints

Next steps