You can use this MCP server to scrape and map websites, run async scrape jobs, and inspect crawlbrulee account usage from MCP-aware agents.
Scrape a URL synchronously: return markdown, cleaned HTML, raw HTML, links, images, metadata, and screenshots.
Control each scrape: JS rendering, proxy tier, country/locale, cleanup selectors, ad/cookie removal, cache age, and screenshot options.
Run long scrapes asynchronously: submit jobs, poll status, fetch results, and optionally receive signed completion webhooks.
Map sites: discover URLs from sitemaps and homepage links, with type filters, pagination, and discovery caps.
Check account state: get billing-cycle credit usage and identify the configured org/token via whoami.
Operate as a zero-install MCP server: run with npx over stdio in Claude Code, Cursor, Codex, Claude Desktop, etc.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@@crawlbrulee/mcp Scrape the product page at https://crawlbrulee.com and return markdown with links"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
EU-native web scraping for AI agents & developers.
plug crawlbrulee into your agent. the official mcp server for crawlbrulee gives mcp-aware agents — Claude Code, Codex, Cursor, Claude Desktop — native tools to scrape pages, map sites, run background jobs, and check usage. one call turns any url into clean markdown, screenshots, metadata and links.
everything runs in the EU. the fetch, the render, the cache and your result never leave EU servers. the proxy exit is the one hop you choose: pick an EU exit and nothing leaves at all. gdpr-aligned, with a data processing agreement.
output made for models. markdown with ads, popups and cookie banners removed and the links kept, ready for the prompt. full-page screenshots can come back as tiles sized for an image model.
the hard parts, handled. headless Chrome when a page needs it, rotating proxies with country selection, automatic retries, ad and cookie-banner removal, caching, background jobs and signed webhooks.
start free. 750 credits, no credit card.
get a free api key → dashboard.crawlbrulee.com
the server:
npx-runnable — zero install.
wraps the @crawlbrulee/sdk under the hood; this mcp is just a thin protocol adapter.
stdio transport for terminal-based agents.
strict, fully-described tool schemas — agents see what every parameter does without reading docs.
this readme covers the mcp server itself — its tools and how to wire it into a host. for how the api behaves — endpoints, parameters, and error semantics — please see our api docs.
# Claude Code
claude mcp add crawlbrulee \
--env CRAWLBRULEE_API_KEY=cwbl_... \
-- npx -y @crawlbrulee/mcp
# Cursor — add to ~/.cursor/mcp.json:
{
"mcpServers": {
"crawlbrulee": {
"command": "npx",
"args": ["-y", "@crawlbrulee/mcp"],
"env": { "CRAWLBRULEE_API_KEY": "cwbl_..." }
}
}
}the same pattern works for Codex, Claude Desktop, and any other host that
accepts a stdio mcp launch command — set command: npx, args: ["-y", "@crawlbrulee/mcp"], and forward CRAWLBRULEE_API_KEY via the env block.
Related MCP server: ZapFetch MCP Server
env var | required | description |
| yes | api key sent as |
the mcp reads the env var on first tool invocation — not at startup — so a typo in your config surfaces as a clear tool-error message rather than the server failing to come up. see authentication for how the api consumes keys.
scrapefetch a single url and return the requested content (markdown, cleaned html, raw html, links, images, screenshot, page metadata).
input — only url is required; everything else has sane defaults.
{
"url": "https://example.com",
"extract": {
"markdown": true,
"links": true,
"screenshot": { "type": "full_page", "device_mode": "desktop" },
},
"require_js": false,
"proxy": "basic",
"cleanup": { "ads_and_popups": true, "exclude_selectors": ["nav", "footer"] },
"cache": { "max_age": 3600 },
"location": { "locale": "en-US", "country": "US" },
"zero_data_retention": false,
}zero_data_retention (boolean, default false) keeps the result out of the shared cache; anything stored to deliver it is kept for 24 hours, then deleted. it adds 1 credit and must be enabled for your organization. scrape_async takes it too. see zero data retention.
named values with extract.elements. to pull a few values from a page — prices, titles, links — give each one a name and a CSS selector, instead of reading the whole markdown. they come back in elements under the same names. all: true returns every match as a list, and fields reads named values inside each match. it costs no extra credits. scrape_async takes it too.
{
"url": "https://books.toscrape.com/",
"extract": {
"elements": {
"heading": "h1",
"books": {
"selector": "article.product_pod",
"all": true,
"fields": {
"title": { "selector": "h3 a", "output": "attribute", "attribute": "title" },
"price": ".price_color",
},
},
},
},
}
// → "elements": {
// "heading": "All products",
// "books": [{ "title": "A Light in the Attic", "price": "£51.77" }, ...]
// }a name with no match is null (or [] with all: true). limits and selector rules: see elements.
output — full scrape result. page metadata (title, OG tags, etc.) is returned under metadata. extracted images are returned as absolute urls — query strings are preserved, and relative srcs are resolved against the page url. screenshots are returned as signed download urls the agent can fetch separately. in rare cases a screenshot can't be captured: when you requested other outputs too, the screenshot field is simply left out while the rest is still returned — but a screenshot-only call that can't deliver errors instead (unsupported_screenshot_output, HTTP 422, when the content type can't be screenshotted) and isn't billed. the result also carries page_status_code and a top-level response_meta.usage block:
{
"url": "https://example.com",
"requested_url": "https://example.com",
"page_status_code": 200, // the site's own HTTP status for the final page
"markdown": "...",
"metadata": { "title": "Example Domain" },
"response_meta": {
"usage": {
// total_credit_cost = engine_credit_cost × proxy_multiplier + screenshot_slicing_credit_cost + zero_data_retention_credit_cost
"total_credit_cost": 1,
"engine_credit_cost": 1, // http 1, browser 3, screenshot 5, cache 0
"proxy_multiplier": 1, // basic 1, advanced 5
"screenshot_slicing_credit_cost": 0, // 1 when the screenshot was split into slices, otherwise 0
"zero_data_retention_credit_cost": 0, // 1 when zero_data_retention added its credit, otherwise 0
"engine": "http", // "http" | "browser" | "screenshot" | "cache"
"proxy": "basic", // resolved tier actually used: "basic" | "advanced" (never "auto")
},
},
}a page the site served is a result, not an error. page_status_code is the HTTP status the site answered with for the final page, after redirects. a 404, 410 or 503 page comes back with its content and its status here, so check page_status_code before you trust the content: a 404 means the markdown is the site's "not found" page. 2xx and 4xx pages are billed, except 403, 407, 408, 429 and 451; 5xx pages are never billed. when the site can't be reached at all, the tool returns a target_unreachable error instead (see errors).
alongside response_meta.usage, the result surfaces any non-fatal warnings — stable string codes an agent can switch on. an outsized page is truncated rather than refused, and the code names which part was cut:
code | what it means for the payload |
| the page was taller than the scrolling-capture height cap; the screenshot covers the top of the page. |
| the page had more than 30,000 links; the |
| the page had more than 10,000 inline images; the |
| the page body exceeded 10,000,000 characters; |
| an |
one more code is not about size: screenshot_unavailable means a screenshot was asked for, but the page came back from the http engine without one. the rest of the result is still there.
and if you requested an extract that doesn't apply to the content type (e.g. metadata of a JSON file), the field name comes back in an unsupported_fields list — with the rest of the payload still returned.
every input field, its default, and its constraints are documented under the scrape endpoint — with extraction, screenshots, proxies & location, and caching covering the individual blocks.
scrape_asyncsubmit a scrape job to run asynchronously and get back a job_id immediately, instead of holding the connection open. use this for long-running scrapes (heavy js rendering, full-page screenshots of long pages); for a quick one-shot fetch prefer the synchronous scrape tool. then poll scrape_status until the job is done and fetch the page with scrape_result.
takes the same input as scrape plus an optional per-job completion webhook:
{
"url": "https://example.com",
"extract": { "markdown": true },
"webhook": {
// Endpoint that receives one signed `scrape.complete` POST when the job
// finishes. http/https (HTTPS required in production), max 2048 chars.
"url": "https://hooks.example.com/cwbl",
// Opaque correlation object echoed back verbatim in the delivery's
// `data.metadata`. Serializes to at most 2048 bytes.
"metadata": { "ref": "order-42" },
},
}output — { "job_id": "..." }.
when a webhook is attached, we deliver a single signed scrape.complete POST to your endpoint once the job reaches a terminal state, with your metadata echoed under data.metadata and the job's usage under data.response_meta.usage — so you can react to completion (and track cost) without polling. verify the X-Cwbl-Signature header with the sdk's verifyWebhookSignature (configure the signing secret in the dashboard under account → webhooks).
the job lifecycle is documented under async scrape; the delivery contract and payload shape under webhooks, with the signature scheme in webhook verification.
scrape_statuslook up the current lifecycle status of an async job: pending, running, done, or failed (with an error message when failed). once the job is done the response also carries a response_meta.usage block (total_credit_cost and its parts, billed engine, resolved proxy tier). a cache hit is represented by engine: "cache". a job whose page the site served ends done even when that page is a 404 — read page_status_code in the result. poll until done, then call scrape_result.
{ "job_id": "..." }scrape_resultfetch the extracted content of a completed async job — the same result shape as the synchronous scrape tool (including page_status_code, metadata, elements and response_meta.usage). errors if the job is still pending/running, so check scrape_status first.
{ "job_id": "..." }mapbuild (or fetch a cached) link-map for a website. combines sitemap discovery with homepage link extraction. use this to enumerate a site before scraping selected pages. each link is just { url }.
max_urls (default 5000, max 100000) is a discovery budget, not a trim at the end — discovery stops as soon as that many urls are found, so a smaller value is a faster, cheaper crawl. limit (default 5000, max 10000) only pages the answer.
returned urls are normalized the same way scrape normalizes its returned url, so map-then-scrape stays on one host. results are ordered with the most useful links first.
the response's response_meta carries pagination, truncation, and a usage block (total_credit_cost = engine_credit_cost × proxy_multiplier + zero_data_retention_credit_cost, billed engine, resolved proxy tier). map has no screenshot slicing and no page_status_code, since it reads many files, not one page. a map that found nothing because the site answered only with statuses we don't bill (a 5xx, for example), or not at all, is empty and free.
{
"url": "https://example.com",
"sitemap_only": false,
"types": { "internal": true, "external": false, "internal_subdomains": true },
"max_urls": 5000,
"page": 1,
"limit": 1000,
"zero_data_retention": false,
}zero_data_retention works as it does on scrape (see zero data retention).
a map stopped by your own max_urls returns exactly that many links with response_capped: false — the signal that the site has more is truncation.discovery_cap_reason:
{
"truncation": {
"storage_capped": false,
"response_capped": false,
"total_before_max_urls": 5000,
"total_detected_before_storage_cap": 5000,
"discovery_capped": true, // discovery stopped before reading every sitemap file
"sitemaps_skipped": 3, // files skipped or only partly read
"discovery_cap_reason": "max_urls", // retry with a higher max_urls
},
}discovery_cap_reason is one of max_urls, time, file_budget, depth, file_size, unread_files, or null when nothing stopped discovery. only max_urls is a limit you can raise from the request. unread_files means a sitemap file the site publishes could not be read at all this time — often temporary, so asking again later can return more. time, file_budget, depth and file_size mean the site itself is big, slow or deep, and a retry will not help.
see the map endpoint for discovery rules and pagination semantics.
usagereturns the current billing-cycle snapshot: total / used / available credits, used quota percent, max concurrency, and cycle reset timestamp. takes no arguments. what a call costs, and how credits are counted, is documented under credits & pricing.
whoamireturns the organization name, token name, and truncated token preview for the configured api key. useful for confirming which account is in use before credit-consuming operations.
every tool returns an mcp error result (isError: true) when the api call fails. the error text follows a stable format:
[<errorName>] <message> (HTTP <status>)a few codes add a short next step after that, e.g. target_unreachable and zero_data_retention_not_enabled. branch on the errorName code, not on the rest of the text.
agents can branch on the errorName code. the set comes from the sdk's ApiErrorName union plus two synthetic codes added by this mcp (missing_api_key, internal_error):
code | meaning |
|
|
| server rejected the api key (revoked, wrong env, etc.). |
| temporary backend failure (HTTP 503). your key is fine — retry with backoff. |
| rate limit hit — back off and retry. |
| plan credit / concurrency cap exceeded. show |
| input failed server validation. |
| target url was rejected before fetching. |
| target url is on the blocklist. |
| origin's anti-bot defenses blocked the fetch. |
| origin redirected the fetch in a loop (HTTP 422). the target's doing — don't retry blindly. |
| the page's html was too large to process (HTTP 422). terminal — never retry it. |
| we could not reach the site at all (HTTP 502). not billed. retry later or check the url. |
|
|
| the scrape could not be completed. a page the site served, even a 404, is never this error. |
| screenshot-only request on a content type that can't be screenshotted (HTTP 422). not billed. |
| async job ID unknown, or submitted more than 24 hours ago ( |
| network / read timeout. safe to retry. |
| caller cancelled before completion. |
| unhandled server-side failure. |
| sdk error without a typed name. |
| bug in this mcp — please open an issue. |
the api docs carry the canonical error reference — every error name, what causes it, and how to recover.
pnpm install
pnpm typecheck # tsc --noEmit
pnpm lint # eslint
pnpm test # vitest run
pnpm build # tsup → dist/index.js with shebang
pnpm verify # all of the aboverun the built mcp locally:
CRAWLBRULEE_API_KEY=cwbl_... node ./dist/index.jsit will block waiting for an mcp client on stdio. combine with the MCP Inspector for interactive debugging.
this readme covers the mcp server itself — installing it, wiring it into a host, and the tools it exposes. for how the api behaves — endpoints, parameters, and error semantics — the api docs are canonical. the mcp guide covers host setup in more depth.
one api, many ways to call it:
js/ts sdk — @crawlbrulee/sdk (the sdk this mcp wraps)
python sdk — crawlbrulee on pypi
cli — npx crawlbrulee
mcp server — @crawlbrulee/mcp (this one)
agent skills — for skills-aware coding agents
docs: crawlbrulee.com/docs · dashboard: dashboard.crawlbrulee.com
Build (or fetch a cached) link-map for a website by combining sitemap discovery with homepage link extraction. Returns paginated lists of discovered URLs, each item just { url }. Filterable by link type (internal / external / subdomain). Use this to enumerate a site before scraping selected pages. max_urls (default 5000, max 100000) is a discovery budget, not a trim at the end: discovery stops as soon as that many URLs are found, so a smaller value is a faster, cheaper crawl. A map stopped that way returns exactly max_urls links with response_capped false — the signal that the site has more is response_meta.truncation.discovery_cap_reason. When that is "max_urls", ask again with a higher max_urls to get more; "unread_files" means a sitemap file could not be read this time and is often temporary, so asking again later can return more; "time", "file_budget", "depth" and "file_size" mean the site itself is big, slow or deep and a retry will not help. discovery_capped says discovery stopped early, sitemaps_skipped how many sitemap files were skipped or only partly read. limit (default 5000, max 10000) only pages the answer. Returned URLs are normalized the same way scrape normalizes its returned url, so map-then-scrape stays on one host. Results are ordered with the most useful links first. The response carries response_meta.usage: total_credit_cost = engine_credit_cost × proxy_multiplier, plus engine and proxy (the resolved tier, never auto); credits is the old, deprecated name of total_credit_cost. Map has no screenshot slicing and no page_status_code: it reads many files, not one page. A map that found nothing because the site answered only with statuses we don't bill (a 5xx, for example), or not at all, is empty and free. If the site cannot be reached, the tool can return a target_unreachable error (HTTP 502, not billed): retry later or check the URL.
| Name | Required | Description | Default |
|---|---|---|---|
| url | Yes | The website URL to map. Mapping always targets the site root, so the path, query string and fragment are dropped; known tracking parameters are removed before the request is processed. | |
| page | No | Page number for paginated results | |
| cache | No | Cache settings for this request | |
| limit | No | Number of URLs to return per page. Default 5000, maximum 10000. | |
| proxy | No | Proxy tier to use for fetching | auto |
| types | No | Filter which link types to include | |
| location | No | Optional country emulation for the map | |
| max_urls | No | Maximum number of URLs to discover and store in the map. Default 5000, maximum 100000. Sitemap discovery stops as soon as this many URLs have been found, so a smaller value is a faster and lighter crawl, not just a smaller answer. | |
| sitemap_only | No | Only use sitemap.xml — skip homepage link extraction |
| Name | Required | Description |
|---|---|---|
| links | Yes | List of discovered URLs for the current page |
| response_meta | Yes | Response metadata including pagination, truncation, and usage info |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full burden and does so thoroughly: billing semantics (credit cost formula, empty-and-free case), error surface (target_unreachable, HTTP 502, not billed), and the crucial distinction that max_urls is a discovery budget rather than a post-hoc trim. This is exactly the behavioral context an agent needs.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Purpose is front-loaded and the dense truncation/retry paragraph is genuinely load-bearing for correct retry behavior. It is long and could be tightened, but almost every sentence adds decision-relevant information rather than filler.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a 9-parameter tool with nested objects and an output schema, the description is complete: it explains the response_meta fields, deprecation of 'credits', pagination, and error states even though the output schema exists. Nothing an agent needs to call this correctly is missing.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100% so the baseline is 3, but the description meaningfully adds: max_urls stops discovery early (faster/cheaper crawl), limit only pages the answer, proxy resolves to a concrete tier (never 'auto'), and map-then-scrape URL normalization matches scrape. Some of this restates the schema, but the discovery-budget semantics go beyond it.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb and resource ('Build (or fetch a cached) link-map for a website') and names the mechanism (sitemap discovery + homepage link extraction). It also routes the agent to a sibling use case: 'enumerate a site before scraping selected pages,' distinguishing it from scrape.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicitly says when to use it (enumerate before scraping) and gives detailed when-to-retry guidance keyed to discovery_cap_reason values, including which reasons a retry will not help. This is unusually complete invocation guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Fetch a single URL via the crawlbrulee scraping API and return the requested content (markdown, cleaned HTML, raw HTML, links, images, screenshot, page metadata in metadata). Use this for one-shot page extraction. For full-site discovery use the map tool first. Check page_status_code before you trust the content: it is the HTTP status the site answered with for the final page. A page the site served is a successful result whatever its status, so a 404 or 503 page comes back with its content, not as an error. A 404 means the markdown is the site's "not found" page, not the page you asked for. 2xx and 4xx pages are billed, except 403, 407, 408, 429 and 451; 5xx pages are never billed. If the site cannot be reached at all the tool returns a target_unreachable error (HTTP 502, not billed): retry later or check the URL. Screenshot URLs in the response are signed download links — the agent can fetch them when needed. The response also carries response_meta.usage: total_credit_cost = engine_credit_cost × proxy_multiplier + screenshot_slicing_credit_cost, plus engine and proxy (the resolved tier — never auto). engine is the billed engine (http, browser, screenshot, cache); a cache hit is represented by engine: "cache". credits and screenshot_slices are the old, deprecated names of total_credit_cost and screenshot_slicing_credit_cost. Extraction is capped per page: 30,000 links, 10,000 inline images and 10,000,000 characters of HTML. A page past a cap is truncated rather than refused, and the response warnings array names which one (links_truncated, inline_images_truncated, raw_html_truncated) — so treat that output as incomplete. Use cleanup to control what is removed before any output is built: ads_and_popups (on by default) drops ads, cookie banners and chat widgets, and exclude_selectors removes anything else. It shapes markdown, cleaned_html, links, images and the screenshot, and NEVER raw_html — so request raw_html when you need the page exactly as it arrived. warnings also reports a section whose extraction failed outright (links_unavailable, inline_images_unavailable, metadata_unavailable): that field comes back omitted or empty while the rest of the scrape succeeded, so do NOT conclude the page had no links/images/metadata — re-run the scrape instead. An empty field with no such warning does mean the page had none.
| Name | Required | Description | Default |
|---|---|---|---|
| url | Yes | The URL to scrape. Known tracking parameters are removed before the page is fetched, so they are neither sent to the target site nor part of the cache key. Every other query parameter is kept verbatim and is part of the cache key. | |
| cache | No | Cache settings for this request | |
| proxy | No | Proxy tier to use for fetching | auto |
| cleanup | No | What is removed from the page before any output is built. Applies to markdown, cleaned_html, links and images on every engine, and to the screenshot. Never applies to raw_html, which is always the page before we removed anything. | |
| extract | No | Which content formats to extract. Defaults to metadata + cleaned_html. | |
| location | No | Optional locale + country emulation for the scrape | |
| require_js | No | Use a headless browser to render JavaScript before scraping |
| Name | Required | Description |
|---|---|---|
| url | Yes | The URL that was actually scraped, after any redirects, in normalized form — the base that links, images, and internal labels are computed against |
| links | No | Links found on the page |
| images | No | Inline images found on the page |
| markdown | No | Page content converted to clean Markdown |
| metadata | No | Extracted page metadata (title, OG tags, etc.) |
| raw_html | No | Raw, unprocessed HTML of the page |
| warnings | No | Non-error notices about the scrape. Truncation codes — `screenshot_truncated` (long page exceeded the scrolling-screenshot height cap), `links_truncated` / `inline_images_truncated` (page had more links/images than the per-page extraction caps), `raw_html_truncated` / `metadata_truncated` (rendered HTML exceeded the per-page size budget) — mean the field is present but capped. Unavailability codes — `links_unavailable` / `inline_images_unavailable` / `metadata_unavailable` — mean that optional field could not be extracted and was omitted (null/empty) while the rest of the scrape succeeded, so an empty field carrying one of these does NOT mean the page had none. Stable string codes — clients can switch on them. Warnings are stored with the result: async result fetches and cache hits carry them too, filtered to the fields the request asked for. |
| screenshot | No | Screenshot of the page, if requested |
| cleaned_html | No | Cleaned HTML of the main page content |
| content_type | No | Content-Type header returned by the server |
| requested_url | Yes | The URL you requested, echoed verbatim — before any redirects |
| response_meta | Yes | Request-level metadata. `response_meta.usage` reports what the request cost (total_credit_cost and its parts), the billed engine and the resolved proxy tier. |
| page_status_code | No | The HTTP status the site answered with for the final page, after redirects. Check it before you trust the content. A page the site served is a successful result whatever its status: a 404 or 503 page comes back with its content here, not as an error, so a 404 means the markdown is the site's "not found" page, not the page you asked for. Billing follows this status: 2xx and 4xx pages are billed, except 403, 407, 408, 429 and 451; 5xx pages are never billed. Older api versions do not send it; they only returned pages that loaded normally. |
| unsupported_fields | No | Extract fields that were requested but are not supported for this content type |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full behavioral burden and does so richly: status-code semantics, what is/isn't billed, truncation caps and their warnings, unavailable-section warnings vs genuinely empty fields, cleanup scope versus raw_html, signed screenshot links, and cache/engine/proxy resolution. This is unusually complete context beyond any structured field.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loaded with purpose then usage then caveats, and nearly every sentence conveys a distinct operational fact (billing, truncation, warnings, retry). It is nonetheless a very long block, and the paragraph of billing/warning detail is dense enough that trimming would improve scanning.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a 7-parameter tool with nested objects, no annotations, and an output schema present, the description supplies everything an agent needs to invoke correctly and interpret results: usage routing, status handling, truncation, billing, error paths, and cleanup scope. Return-value explanation is appropriately left to the output schema.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, so the baseline is 3, but the description adds genuine meaning beyond the schema: the interaction rule that cleanup shapes markdown/cleaned_html/links/images/screenshot but NEVER raw_html, and that truncation is announced via named warnings rather than by refusing the request. It does not add format detail for cache, proxy, location, or require_js, which the schema already covers.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Opens with a specific verb (Fetch) plus resource (single URL), enumerates the returnable content formats, and explicitly distinguishes itself from the `map` sibling for full-site discovery. An agent can tell this is the one-shot page extractor without opening the schema.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Gives a clear usage context ('Use this for one-shot page extraction') and names the alternative (`map`) with the condition that selects it, plus retry guidance for the target_unreachable error. It does not, however, differentiate itself from the scrape_async/scrape_status/scrape_result siblings, which an agent would also need to choose among.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Submit a scrape job to run ASYNCHRONOUSLY and return a job_id immediately, instead of holding the connection open. Use this for long-running scrapes (heavy JS rendering, full-page screenshots of long pages) — for a quick one-shot fetch prefer the synchronous scrape tool, which blocks and returns the page directly. Poll the job with scrape_status and fetch the page with scrape_result once done. Optionally attach a per-job completion webhook: we deliver a single signed scrape.complete POST to your endpoint when the job finishes (HTTPS required in production), and echoes your opaque webhook.metadata back in the delivery (under data.metadata, alongside data.page_status_code (on success) and data.response_meta.usage).
| Name | Required | Description | Default |
|---|---|---|---|
| url | Yes | The URL to scrape. Known tracking parameters are removed before the page is fetched, so they are neither sent to the target site nor part of the cache key. Every other query parameter is kept verbatim and is part of the cache key. | |
| cache | No | Cache settings for this request | |
| proxy | No | Proxy tier to use for fetching | auto |
| cleanup | No | What is removed from the page before any output is built. Applies to markdown, cleaned_html, links and images on every engine, and to the screenshot. Never applies to raw_html, which is always the page before we removed anything. | |
| extract | No | Which content formats to extract. Defaults to metadata + cleaned_html. | |
| webhook | No | Completion webhook delivered when this async job finishes. Async-only: the synchronous `scrape` tool does not accept it. | |
| location | No | Optional locale + country emulation for the scrape | |
| require_js | No | Use a headless browser to render JavaScript before scraping |
| Name | Required | Description |
|---|---|---|
| job_id | Yes | Job identifier — pass it to the `scrape_status` / `scrape_result` tools |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden and it does so well: it discloses non-blocking return of a job_id, webhook delivery semantics (single signed scrape.complete POST, HTTPS required in production), and metadata echoing. It stops short of error/retry behavior or concurrency limits, which keeps it from a 5.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The key async-vs-sync distinction and the job_id return are front-loaded, then alternatives and follow-up tools, then webhook detail. The trailing webhook payload digression is dense and partly redundant with the schema, but each clause is informative rather than filler.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
An output schema exists, so return values need not be re-explained, and the description correctly covers the async lifecycle end to end. It leaves some operational details unstated (timeouts, failure delivery behavior), but nothing an agent needs to submit the job correctly is missing.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the schema already documents url, cache, proxy, cleanup, extract, webhook and location in depth. The description's webhook prose largely restates what the schema says, adding only marginal framing rather than new parameter meaning; the baseline 3 applies.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb and resource (submit a scrape job) plus the immediate return value (job_id) and the core behavioral distinction from the sync path. An agent can tell exactly what this does and how it differs from the sibling `scrape` without reading any schema.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicitly says when to use this (long-running scrapes, heavy JS rendering, full-page screenshots) and when not to (quick one-shot fetch should use synchronous `scrape`). It also names the lifecycle partners `scrape_status` (poll) and `scrape_result` (fetch), routing the agent through the whole workflow.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Fetch the extracted content of a completed async scrape job (the same result shape as the synchronous scrape tool: markdown, cleaned HTML, raw HTML, links, images, screenshot, page metadata in metadata, and response_meta.usage). Errors if the job is still pending/running — check scrape_status first (status done) before calling this. Check page_status_code before you trust the content: it is the HTTP status the site answered with for the final page. A page the site served is a successful result whatever its status, so a 404 or 503 page comes back with its content, not as an error. A 404 means the markdown is the site's "not found" page, not the page you asked for. 2xx and 4xx pages are billed, except 403, 407, 408, 429 and 451; 5xx pages are never billed. Screenshot URLs are signed download links.
| Name | Required | Description | Default |
|---|---|---|---|
| job_id | Yes | Job identifier returned by the `scrape_async` tool |
| Name | Required | Description |
|---|---|---|
| url | Yes | The URL that was actually scraped, after any redirects, in normalized form — the base that links, images, and internal labels are computed against |
| links | No | Links found on the page |
| images | No | Inline images found on the page |
| markdown | No | Page content converted to clean Markdown |
| metadata | No | Extracted page metadata (title, OG tags, etc.) |
| raw_html | No | Raw, unprocessed HTML of the page |
| warnings | No | Non-error notices about the scrape. Truncation codes — `screenshot_truncated` (long page exceeded the scrolling-screenshot height cap), `links_truncated` / `inline_images_truncated` (page had more links/images than the per-page extraction caps), `raw_html_truncated` / `metadata_truncated` (rendered HTML exceeded the per-page size budget) — mean the field is present but capped. Unavailability codes — `links_unavailable` / `inline_images_unavailable` / `metadata_unavailable` — mean that optional field could not be extracted and was omitted (null/empty) while the rest of the scrape succeeded, so an empty field carrying one of these does NOT mean the page had none. Stable string codes — clients can switch on them. Warnings are stored with the result: async result fetches and cache hits carry them too, filtered to the fields the request asked for. |
| screenshot | No | Screenshot of the page, if requested |
| cleaned_html | No | Cleaned HTML of the main page content |
| content_type | No | Content-Type header returned by the server |
| requested_url | Yes | The URL you requested, echoed verbatim — before any redirects |
| response_meta | Yes | Request-level metadata. `response_meta.usage` reports what the request cost (total_credit_cost and its parts), the billed engine and the resolved proxy tier. |
| page_status_code | No | The HTTP status the site answered with for the final page, after redirects. Check it before you trust the content. A page the site served is a successful result whatever its status: a 404 or 503 page comes back with its content here, not as an error, so a 404 means the markdown is the site's "not found" page, not the page you asked for. Billing follows this status: 2xx and 4xx pages are billed, except 403, 407, 408, 429 and 451; 5xx pages are never billed. Older api versions do not send it; they only returned pages that loaded normally. |
| unsupported_fields | No | Extract fields that were requested but are not supported for this content type |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full burden and does so richly: error behavior for pending/running jobs, the critical insight that non-2xx pages return content rather than errors, the meaning of `page_status_code`, detailed billing rules (403/407/408/429/451 exceptions, 5xx never billed), and signed screenshot URLs. This is far beyond what the schema provides.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loaded with the core verb/resource and the result shape, with prerequisites and status-code caveats following in a logical order. It is dense and somewhat long, but nearly every sentence conveys actionable behavioral information rather than filler.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Despite an output schema being present (so return values need not be explained), the description still usefully maps the result shape and adds the prerequisites, error conditions, and billing semantics an agent needs. Nothing required to invoke and interpret this tool correctly is missing.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Only one parameter with 100% schema description coverage, where the schema already documents `job_id` as the identifier from `scrape_async`. The description adds no syntax or format detail beyond what the schema already states, so the baseline 3 applies.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb (fetch) and resource (extracted content of a completed async scrape job) and ties the result shape to the sibling `scrape` tool. An agent can immediately tell it apart from `scrape_async` (starts the job) and `scrape_status` (checks progress).
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicitly states the precondition (job must be done) and names the alternative to check first: 'check `scrape_status` first (status `done`) before calling this.' It also names the failure mode (errors if pending/running), so the agent knows exactly when and when-not to call.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Look up the current lifecycle status of an async scrape job submitted via scrape_async. Returns the job state (pending, running, done, failed), with an error message when it failed and a response_meta.usage block (total_credit_cost and its parts, billed engine, resolved proxy tier) once it is done. A job whose page the site served ends done even when that page is a 404 or 503 — read page_status_code in the result. A job that could not reach the site at all ends failed (the error text is a general message, not the reason) and is not billed; retry later or check the url. Poll this until the status is done, then call scrape_result to fetch the page. If you registered a completion webhook on submit you can skip polling and react to the delivery instead.
| Name | Required | Description | Default |
|---|---|---|---|
| job_id | Yes | Job identifier returned by the `scrape_async` tool |
| Name | Required | Description |
|---|---|---|
| error | No | Error message if the job ended in `failed`. A job whose page the site served, even a 404 page, ends `done`, not `failed`: read `page_status_code` in the `scrape_result`. |
| job_id | Yes | The job identifier |
| status | Yes | Current state of the job (pending, running, done, failed) |
| created_at | Yes | ISO-8601 UTC timestamp when the job was created |
| response_meta | No | Usage accounting for the finished job. Present only once the job is `done`; `response_meta.usage` reports what the job cost (total_credit_cost and its parts), the billed engine and the resolved proxy tier. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations exist, so the description carries the full burden and does so richly: it enumerates the states, explains that a served 404/503 still ends `done` (read page_status_code), that an unreachable site ends `failed` and is not billed, that the error text is generic, and that the usage/billing block appears only once done. These are real behavioral traits an agent needs, not restatements.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The primary action and the poll-then-fetch flow are front-loaded, followed by edge-case semantics and the webhook alternative. Despite being a paragraph, every sentence delivers distinct operational information (state meanings, billing, retry, next call) with no filler.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
An output schema exists, yet the description still surfaces the result fields an agent must interpret (page_status_code, error, response_meta.usage) and couples them to next actions. Nothing needed to poll, interpret, or act on the job state is missing.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Single parameter with 100% schema description coverage; the schema already defines job_id as the identifier returned by `scrape_async`. The description repeats that origin but adds no format, syntax, or validity detail beyond the schema, so the baseline 3 applies.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb (look up) and resource (lifecycle status of an async scrape job) and immediately names the sibling that creates the job (`scrape_async`) plus the follow-up tool (`scrape_result`). An agent can distinguish this from scrape, scrape_async, and scrape_result without opening any schema.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicitly gives when-to-use and the follow-up: 'Poll this until the status is `done`, then call `scrape_result`.' It also names the alternative path ('if you registered a completion webhook on submit you can skip polling') and the failure path ('retry later or check the url'), so the routing decision is fully specified.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Return the current billing-cycle usage for the authenticated organization: total / used / available credits, used quota percent, max concurrency, and cycle reset time. Call this before launching large scrape jobs to confirm available credit.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
| usage_reset | Yes | ISO 8601 UTC timestamp when the current billing cycle ends and used_credits resets. |
| used_credits | Yes | Credits spent so far in the current billing cycle. May exceed total_credits on plans that allow overages. |
| total_credits | Yes | Total credits available for the current billing cycle (plan base + purchased + gifted). |
| max_concurrency | Yes | Maximum number of concurrent jobs allowed for this organization (plan base + purchased + gifted extras). |
| available_credits | Yes | Remaining credits (max(0, total_credits - used_credits)). Clamped to 0 when in overage. |
| used_quota_percent | Yes | Percentage of total_credits used in the current cycle, rounded to 1 decimal. Not capped — values >100 indicate overage. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full behavioral burden. It clearly states that the tool returns usage data for the authenticated organization, lists the specific metrics, and implies a read-only operation. It does not discuss rate limits or authentication requirements, but the described behavior is transparent.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences, front-loads the return content, and closes with a practical usage directive. Every sentence earns its place with no redundancy.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given zero parameters, an output schema that documents return values, and a straightforward read operation, the description covers what the tool does, what it returns, and when to use it. Nothing essential is missing for an agent to invoke it correctly.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has zero parameters, so the schema is trivially complete. The description adds no parameter-level detail because none is needed; the baseline of 4 applies.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description uses a specific verb ('Return') and names the exact resource ('current billing-cycle usage for the authenticated organization') with a detailed list of returned fields. It is clearly distinct from the sibling scraping tools and whoami.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description explicitly tells the agent when to call this tool ('before launching large scrape jobs to confirm available credit'). It does not explicitly list exclusions or alternatives, but the context is clear enough for correct selection.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Return the organization name, token name, and truncated token preview for the API key currently configured on the MCP server. Useful for confirming which account is in use before performing credit-consuming operations.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
| token_name | Yes | User-assigned name of the API token authenticating the request. |
| token_preview | Yes | Truncated preview of the API token (e.g. `cwbl_…xyz`). Safe to display; does not authenticate. |
| organization_name | Yes | Display name of the organization that owns the token. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It explicitly states the operation is a return (read-only) and discloses that the token preview is truncated, which is a useful security-related behavioral trait. It does not explicitly mention side effects or credit consumption, but for a whoami-style identity check the behavior is simple and adequately conveyed.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two sentences with no wasted words. The first sentence front-loads the action and return values; the second adds a practical usage note. Every word earns its place.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a zero-parameter tool with an output schema, the description is fully complete. It states what information is returned and provides a clear reason to call the tool. No additional context is needed for correct invocation.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool has zero parameters, so the baseline is 4. The description adds no parameter detail, but none is needed since the input schema is empty and there is nothing to explain.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description starts with the verb 'Return' and specifies a concrete resource: the organization name, token name, and truncated token preview for the currently configured API key. This clearly identifies the tool's purpose and differentiates it from the scraping, mapping, and usage sibling tools without ambiguity.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description gives an explicit when-to-use situation: confirming which account is in use before credit-consuming operations. It does not name alternative tools or state when not to use it, but the siblings are so distinct that the context is sufficient for an agent to route correctly.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v1.1.0map7 fields changedOutput schema / properties / response_meta / properties / usage / descriptionPrevious value: -"Usage accounting for this map request: credits charged, the billed engine, and the resolved proxy tier."New value: +"What this map request cost: total_credit_cost = engine_credit_cost × proxy_multiplier, plus the billed engine and the resolved proxy tier."Output schema / properties / response_meta / properties / usage / properties / credits / deprecatedAdded value: +trueOutput schema / properties / response_meta / properties / usage / properties / credits / descriptionPrevious value: -"Credits charged for this map request"New value: +"Deprecated: use total_credit_cost, which always has the same value. This field will be removed in a future version."Output schema / properties / response_meta / properties / usage / properties / engine_credit_costAdded value: +{
+ "description": "The engine base charged, before the proxy multiplier: 1 for \"http\", 0 for \"cache\". Also 0 for an empty map that is not billed. Older api versions do not send it.",
+ "maximum": 9007199254740991,
+ "minimum": 0,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / properties / proxy_multiplierAdded value: +{
+ "description": "The multiplier of the proxy tier the request ran on: 1 for \"basic\", 5 for \"advanced\". Always reported, even when the engine cost is 0. Older api versions do not send it.",
+ "exclusiveMinimum": 0,
+ "maximum": 9007199254740991,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / properties / total_credit_costAdded value: +{
+ "description": "Credits charged for this map request. Always equals engine_credit_cost × proxy_multiplier. 0 for a cache hit, and for an empty map when the site answered only with statuses we do not bill (a 5xx, for example) or not at all. Older api versions do not send it; read `credits` then, which has the same value.",
+ "maximum": 9007199254740991,
+ "minimum": 0,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / requiredPrevious value: -[
- "credits",
- "engine",
- "proxy"
-]New value: +[
+ "engine",
+ "proxy"
+]scrape14 fields changedOutput schema / properties / page_status_codeAdded value: +{
+ "description": "The HTTP status the site answered with for the final page, after redirects. Check it before you trust the content. A page the site served is a successful result whatever its status: a 404 or 503 page comes back with its content here, not as an error, so a 404 means the markdown is the site's \"not found\" page, not the page you asked for. Billing follows this status: 2xx and 4xx pages are billed, except 403, 407, 408, 429 and 451; 5xx pages are never billed. Older api versions do not send it; they only returned pages that loaded normally.",
+ "maximum": 9007199254740991,
+ "minimum": -9007199254740991,
+ "type": "integer"
+}Output schema / properties / response_meta / descriptionPrevious value: -"Request-level metadata. `response_meta.usage` reports credits charged, the billed engine, the resolved proxy tier, and any screenshot-slice add-on."New value: +"Request-level metadata. `response_meta.usage` reports what the request cost (total_credit_cost and its parts), the billed engine and the resolved proxy tier."Output schema / properties / response_meta / properties / usage / descriptionPrevious value: -"Usage accounting for this request: credits charged, billed engine, resolved proxy tier, and screenshot-slice add-on."New value: +"What this request cost and why: total_credit_cost = engine_credit_cost × proxy_multiplier + screenshot_slicing_credit_cost, plus the billed engine and the resolved proxy tier."Output schema / properties / response_meta / properties / usage / properties / credits / deprecatedAdded value: +trueOutput schema / properties / response_meta / properties / usage / properties / credits / descriptionPrevious value: -"Credits actually charged for this request. Always equals the engine base × the proxy multiplier, plus screenshot_slices. A cache hit has engine \"cache\" (base 0), so it costs only the parts we still had to compute — a newly produced screenshot-slice variant — and is otherwise free."New value: +"Deprecated: use total_credit_cost, which always has the same value. This field will be removed in a future version."Output schema / properties / response_meta / properties / usage / properties / engine / descriptionPrevious value: -"The engine base the request was billed at — \"http\" (1 credit), \"browser\" (3), \"screenshot\" (5), or \"cache\" (0, the result was served from cache) — before the proxy multiplier. Reflects what was delivered, never what was requested."New value: +"The engine the request was billed at — \"http\", \"browser\", \"screenshot\", or \"cache\" (the result was served from cache). Reflects what was delivered, never what was requested."Output schema / properties / response_meta / properties / usage / properties / engine_credit_costAdded value: +{
+ "description": "The engine base charged, before the proxy multiplier: 1 for \"http\", 3 for \"browser\", 5 for \"screenshot\", 0 for \"cache\". Also 0 when the page is not billed (see page_status_code). Older api versions do not send it.",
+ "maximum": 9007199254740991,
+ "minimum": 0,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / properties / proxy / descriptionPrevious value: -"The proxy tier the request actually ran on (resolved value — never `auto`; `auto` is resolved server-side to `basic` or `advanced`)."New value: +"The proxy tier the request actually ran on (resolved value — never `auto`; `auto` is resolved server-side to `basic` or `advanced`). \"advanced\" multiplies the engine base by 5."Output schema / properties / response_meta / properties / usage / properties / proxy_multiplierAdded value: +{
+ "description": "The multiplier of the proxy tier the request ran on: 1 for \"basic\", 5 for \"advanced\". Always reported, even when the engine cost is 0. Older api versions do not send it.",
+ "exclusiveMinimum": 0,
+ "maximum": 9007199254740991,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / properties / screenshot_slices / deprecatedAdded value: +trueOutput schema / properties / response_meta / properties / usage / properties / screenshot_slices / descriptionPrevious value: -"Whether this request was billed the screenshot-split add-on: 1 when the screenshot was split on this request (a flat +1 credit outside the proxy multiplier, regardless of how many slices the split produced), else 0 — including a cache hit that reused an already-existing slice variant."New value: +"Deprecated: use screenshot_slicing_credit_cost, which always has the same value. Despite its name this is a 0/1 charge, not a count of slices. This field will be removed in a future version."Output schema / properties / response_meta / properties / usage / properties / screenshot_slicing_credit_costAdded value: +{
+ "description": "The screenshot slicing add-on: 1 when the screenshot was split into slices on this request (a flat +1 credit outside the proxy multiplier, however many slices it made), else 0. A cache hit that reused slices that already existed costs 0. Older api versions do not send it; read `screenshot_slices` then, which has the same value.",
+ "maximum": 9007199254740991,
+ "minimum": 0,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / properties / total_credit_costAdded value: +{
+ "description": "Credits charged for this request. Always equals engine_credit_cost × proxy_multiplier + screenshot_slicing_credit_cost. 0 when nothing is billed: a cache hit, or a page whose status is not billed (see page_status_code). Older api versions do not send it; read `credits` then, which has the same value.",
+ "maximum": 9007199254740991,
+ "minimum": 0,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / requiredPrevious value: -[
- "credits",
- "engine",
- "proxy",
- "screenshot_slices"
-]New value: +[
+ "engine",
+ "proxy"
+]scrape_result14 fields changedOutput schema / properties / page_status_codeAdded value: +{
+ "description": "The HTTP status the site answered with for the final page, after redirects. Check it before you trust the content. A page the site served is a successful result whatever its status: a 404 or 503 page comes back with its content here, not as an error, so a 404 means the markdown is the site's \"not found\" page, not the page you asked for. Billing follows this status: 2xx and 4xx pages are billed, except 403, 407, 408, 429 and 451; 5xx pages are never billed. Older api versions do not send it; they only returned pages that loaded normally.",
+ "maximum": 9007199254740991,
+ "minimum": -9007199254740991,
+ "type": "integer"
+}Output schema / properties / response_meta / descriptionPrevious value: -"Request-level metadata. `response_meta.usage` reports credits charged, the billed engine, the resolved proxy tier, and any screenshot-slice add-on."New value: +"Request-level metadata. `response_meta.usage` reports what the request cost (total_credit_cost and its parts), the billed engine and the resolved proxy tier."Output schema / properties / response_meta / properties / usage / descriptionPrevious value: -"Usage accounting for this request: credits charged, billed engine, resolved proxy tier, and screenshot-slice add-on."New value: +"What this request cost and why: total_credit_cost = engine_credit_cost × proxy_multiplier + screenshot_slicing_credit_cost, plus the billed engine and the resolved proxy tier."Output schema / properties / response_meta / properties / usage / properties / credits / deprecatedAdded value: +trueOutput schema / properties / response_meta / properties / usage / properties / credits / descriptionPrevious value: -"Credits actually charged for this request. Always equals the engine base × the proxy multiplier, plus screenshot_slices. A cache hit has engine \"cache\" (base 0), so it costs only the parts we still had to compute — a newly produced screenshot-slice variant — and is otherwise free."New value: +"Deprecated: use total_credit_cost, which always has the same value. This field will be removed in a future version."Output schema / properties / response_meta / properties / usage / properties / engine / descriptionPrevious value: -"The engine base the request was billed at — \"http\" (1 credit), \"browser\" (3), \"screenshot\" (5), or \"cache\" (0, the result was served from cache) — before the proxy multiplier. Reflects what was delivered, never what was requested."New value: +"The engine the request was billed at — \"http\", \"browser\", \"screenshot\", or \"cache\" (the result was served from cache). Reflects what was delivered, never what was requested."Output schema / properties / response_meta / properties / usage / properties / engine_credit_costAdded value: +{
+ "description": "The engine base charged, before the proxy multiplier: 1 for \"http\", 3 for \"browser\", 5 for \"screenshot\", 0 for \"cache\". Also 0 when the page is not billed (see page_status_code). Older api versions do not send it.",
+ "maximum": 9007199254740991,
+ "minimum": 0,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / properties / proxy / descriptionPrevious value: -"The proxy tier the request actually ran on (resolved value — never `auto`; `auto` is resolved server-side to `basic` or `advanced`)."New value: +"The proxy tier the request actually ran on (resolved value — never `auto`; `auto` is resolved server-side to `basic` or `advanced`). \"advanced\" multiplies the engine base by 5."Output schema / properties / response_meta / properties / usage / properties / proxy_multiplierAdded value: +{
+ "description": "The multiplier of the proxy tier the request ran on: 1 for \"basic\", 5 for \"advanced\". Always reported, even when the engine cost is 0. Older api versions do not send it.",
+ "exclusiveMinimum": 0,
+ "maximum": 9007199254740991,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / properties / screenshot_slices / deprecatedAdded value: +trueOutput schema / properties / response_meta / properties / usage / properties / screenshot_slices / descriptionPrevious value: -"Whether this request was billed the screenshot-split add-on: 1 when the screenshot was split on this request (a flat +1 credit outside the proxy multiplier, regardless of how many slices the split produced), else 0 — including a cache hit that reused an already-existing slice variant."New value: +"Deprecated: use screenshot_slicing_credit_cost, which always has the same value. Despite its name this is a 0/1 charge, not a count of slices. This field will be removed in a future version."Output schema / properties / response_meta / properties / usage / properties / screenshot_slicing_credit_costAdded value: +{
+ "description": "The screenshot slicing add-on: 1 when the screenshot was split into slices on this request (a flat +1 credit outside the proxy multiplier, however many slices it made), else 0. A cache hit that reused slices that already existed costs 0. Older api versions do not send it; read `screenshot_slices` then, which has the same value.",
+ "maximum": 9007199254740991,
+ "minimum": 0,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / properties / total_credit_costAdded value: +{
+ "description": "Credits charged for this request. Always equals engine_credit_cost × proxy_multiplier + screenshot_slicing_credit_cost. 0 when nothing is billed: a cache hit, or a page whose status is not billed (see page_status_code). Older api versions do not send it; read `credits` then, which has the same value.",
+ "maximum": 9007199254740991,
+ "minimum": 0,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / requiredPrevious value: -[
- "credits",
- "engine",
- "proxy",
- "screenshot_slices"
-]New value: +[
+ "engine",
+ "proxy"
+]scrape_status14 fields changedOutput schema / properties / error / descriptionPrevious value: -"Error message if the job ended in `failed`"New value: +"Error message if the job ended in `failed`. A job whose page the site served, even a 404 page, ends `done`, not `failed`: read `page_status_code` in the `scrape_result`."Output schema / properties / response_meta / descriptionPrevious value: -"Usage accounting for the finished job. Present only once the job is `done`; `response_meta.usage` reports credits charged, the billed engine, the resolved proxy tier, and any screenshot-slice add-on."New value: +"Usage accounting for the finished job. Present only once the job is `done`; `response_meta.usage` reports what the job cost (total_credit_cost and its parts), the billed engine and the resolved proxy tier."Output schema / properties / response_meta / properties / usage / descriptionPrevious value: -"Usage accounting for this request: credits charged, billed engine, resolved proxy tier, and screenshot-slice add-on."New value: +"What this request cost and why: total_credit_cost = engine_credit_cost × proxy_multiplier + screenshot_slicing_credit_cost, plus the billed engine and the resolved proxy tier."Output schema / properties / response_meta / properties / usage / properties / credits / deprecatedAdded value: +trueOutput schema / properties / response_meta / properties / usage / properties / credits / descriptionPrevious value: -"Credits actually charged for this request. Always equals the engine base × the proxy multiplier, plus screenshot_slices. A cache hit has engine \"cache\" (base 0), so it costs only the parts we still had to compute — a newly produced screenshot-slice variant — and is otherwise free."New value: +"Deprecated: use total_credit_cost, which always has the same value. This field will be removed in a future version."Output schema / properties / response_meta / properties / usage / properties / engine / descriptionPrevious value: -"The engine base the request was billed at — \"http\" (1 credit), \"browser\" (3), \"screenshot\" (5), or \"cache\" (0, the result was served from cache) — before the proxy multiplier. Reflects what was delivered, never what was requested."New value: +"The engine the request was billed at — \"http\", \"browser\", \"screenshot\", or \"cache\" (the result was served from cache). Reflects what was delivered, never what was requested."Output schema / properties / response_meta / properties / usage / properties / engine_credit_costAdded value: +{
+ "description": "The engine base charged, before the proxy multiplier: 1 for \"http\", 3 for \"browser\", 5 for \"screenshot\", 0 for \"cache\". Also 0 when the page is not billed (see page_status_code). Older api versions do not send it.",
+ "maximum": 9007199254740991,
+ "minimum": 0,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / properties / proxy / descriptionPrevious value: -"The proxy tier the request actually ran on (resolved value — never `auto`; `auto` is resolved server-side to `basic` or `advanced`)."New value: +"The proxy tier the request actually ran on (resolved value — never `auto`; `auto` is resolved server-side to `basic` or `advanced`). \"advanced\" multiplies the engine base by 5."Output schema / properties / response_meta / properties / usage / properties / proxy_multiplierAdded value: +{
+ "description": "The multiplier of the proxy tier the request ran on: 1 for \"basic\", 5 for \"advanced\". Always reported, even when the engine cost is 0. Older api versions do not send it.",
+ "exclusiveMinimum": 0,
+ "maximum": 9007199254740991,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / properties / screenshot_slices / deprecatedAdded value: +trueOutput schema / properties / response_meta / properties / usage / properties / screenshot_slices / descriptionPrevious value: -"Whether this request was billed the screenshot-split add-on: 1 when the screenshot was split on this request (a flat +1 credit outside the proxy multiplier, regardless of how many slices the split produced), else 0 — including a cache hit that reused an already-existing slice variant."New value: +"Deprecated: use screenshot_slicing_credit_cost, which always has the same value. Despite its name this is a 0/1 charge, not a count of slices. This field will be removed in a future version."Output schema / properties / response_meta / properties / usage / properties / screenshot_slicing_credit_costAdded value: +{
+ "description": "The screenshot slicing add-on: 1 when the screenshot was split into slices on this request (a flat +1 credit outside the proxy multiplier, however many slices it made), else 0. A cache hit that reused slices that already existed costs 0. Older api versions do not send it; read `screenshot_slices` then, which has the same value.",
+ "maximum": 9007199254740991,
+ "minimum": 0,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / properties / total_credit_costAdded value: +{
+ "description": "Credits charged for this request. Always equals engine_credit_cost × proxy_multiplier + screenshot_slicing_credit_cost. 0 when nothing is billed: a cache hit, or a page whose status is not billed (see page_status_code). Older api versions do not send it; read `credits` then, which has the same value.",
+ "maximum": 9007199254740991,
+ "minimum": 0,
+ "type": "integer"
+}Output schema / properties / response_meta / properties / usage / requiredPrevious value: -[
- "credits",
- "engine",
- "proxy",
- "screenshot_slices"
-]New value: +[
+ "engine",
+ "proxy"
+]v1.0.3mapscrapescrape_asyncscrape_resultscrape_statususagewhoamiScored across 7 tools
Each tool has a clearly distinct purpose: scrape (sync), scrape_async (async submit), scrape_status (poll), scrape_result (fetch result), map (site discovery), whoami (account identity), and usage (billing). The async pipeline tools are cleanly separated by their lifecycle stage, and no two tools could be reasonably confused.
All names use snake_case and the scrape family shares a consistent prefix (scrape, scrape_async, scrape_status, scrape_result). Minor deviation exists because some tools are noun-like (usage, status) while others are verb-like (scrape, map), but overall the pattern is predictable.
Seven tools is well-scoped for a scraping/crawling API, covering synchronous scraping, asynchronous job management, site mapping, and account/billing introspection. Each tool earns its place and the set is neither bloated nor thin.
The surface covers the core lifecycle: single-page scrape, async submit/poll/fetch, site map, and account usage. A minor gap is the lack of a batch or whole-site crawl operation that would let an agent scrape multiple discovered URLs in one call, though map-then-scrape workflows are workable.
Web scraping, extraction and search for AI agents. Hosted MCP with workspace API keys.
Web scraping for agents. Point it at a URL and it returns the page as clean markdown, JavaScript-rendered pages included. Point it at a site and it maps the URLs or crawls the section you need in the background, a few pages at a time so results fit in the conversation. Search the web and read full pages, extract fields with a JSON schema you define (validated, never invented), read a store's catalogue or a blog's posts from the platform's own feed, and check whether a page has changed. Failed requests cost nothing. The free plan includes 1,500 credits a month.
Cloud scraping & crawling API for AI agents. Turn any URL into clean, LLM-ready markdown.
Scrape, crawl and search the web for AI agents via MCP.
