An MCP interface into the Bright Data toolset for web scraping, search engine access, and browser automation
The server defines 18 tools across search, scraping, and browser automation. Most tools have descriptions and Zod-based schemas with proper types. However, there are significant gaps: (1) Many descriptions are moderate-length but lack specificity about WHEN to use a tool versus alternatives, e.g., search_engine_batch vs search_engine are not clearly differentiated. (2) Parameter descriptions vary in quality; some are terse ('The URL to scrape') while others explain behavior well. (3) Output schemas are not explicitly documented, we can infer JSON or Markdown returns from descriptions, but no formal response schema is declared. (4) No error guidance for common failure modes (invalid URL, timeout, rate limit). (5) Browser session tools (tools 6 - 18) form a stateful workflow (navigate → click → type → screenshot) but lack clear composition guidance. (6) Tool names are clear action verbs (navigate, click, scroll), but batch variants use inconsistent naming (search_engine_batch vs scrape_batch). The server does NOT expose API keys in parameters (good: API_TOKEN is injected via env). Most tools are READ_ONLY; browser mutations (click, type) are marked WRITE, which is appropriate. Overall, the definitions meet a baseline but fall short of production-grade clarity and composability.
Scrape a single webpage URL with advanced options for content extraction and get back the results in HTML. This tool can unlock any webpage even if it uses bot detection or CAPTCHA.
Scrape a single webpage URL with advanced options for content extraction and get back the results in MarkDown language. This tool can unlock any webpage even if it uses bot detection or CAPTCHA.
Scrape multiple webpages URLs with advanced options for content extraction and get back the results in MarkDown language. This tool can unlock any webpage even if it uses bot detection or CAPTCHA.
Instructions for activating the scraping browser
Click on an element. Avoid calling this unless you know the element selector (you can use other tools to find those)
Output schemas not documented. No formal specification of response fields, types, or structure. Descriptions infer Markdown or JSON returns, but LLMs cannot plan downstream calls without knowing exact response shape.
Batch tool naming ambiguous. 'search_engine_batch' and 'scrape_batch' do not clearly distinguish when to use batch vs single variant. No WHEN guidance in descriptions (e.g., 'Use this for 2+ queries to save round-trips').
No error recovery guidance. Tools lack descriptions of failure modes (invalid URL, timeout, rate limit, zone not found). Errors throw UserError but do not suggest next steps (e.g., 'Try a different URL' or 'Check API_TOKEN is set').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 61 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Get the HTML content of the current page. Avoid using the full_page option unless it is important to see things like script tags since this can be large
Get the text content of the current page
Go back to the previous page
Go forward to the next page
Get all links on the current page, text and selectors It's strongly recommended that you call the links tool to check that your click target is valid
Navigate a scraping browser session to a new URL
Take a screenshot of the current page
Scroll to the bottom of the current page
Scroll to a specific element on the page
Type text into an element
Wait for an element to be visible on the page
Scrape search results from Google, Bing or Yandex. Returns SERP results in markdown (URL, title, description)
Run multiple search queries simultaneously. Returns JSON for Google, Markdown for Bing/Yandex.
Browser session tools form implicit stateful workflow (navigate → click → type) but lack composition guidance. No tool description explains the required sequence or dependencies (e.g., 'Call scraping_browser_navigate first to set up a page context').
Parameter descriptions are terse and do not explain constraints. E.g., 'CSS selector for the element to click' does not explain valid CSS selector syntax, whether to use #id vs .class, or error handling if selector matches zero or multiple elements.
search_engine and search_engine_batch accept 'cursor' for pagination but output structure and cursor format are undocumented. LLM cannot determine if cursor is opaque token or page number, or how to interpret next_cursor from response.
Tool 'scraping_browser_activation_instructions' is a meta-tool (returns setup text) with vague name and minimal description. It does not fit the action-verb pattern and its utility is unclear, should be internal setup documentation, not a user-facing tool.