> ## Documentation Index
> Fetch the complete documentation index at: https://deepline.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Use Firecrawl in Deepline

> Web scraping, crawling, search, and AI extraction. Includes SDK V2 guidance, input constraints, response fields, and Deepline credit cost.

[Human guide: work with your agent →](https://deepline.com/docs/human/providers/firecrawl/guide.md)
Inspect one Firecrawl action before starting provider-backed work. Your first result is a verified input and output contract.

## Provider-specific requirements

### Action selection

* **Single page** → `firecrawl_scrape`. Returns markdown by default.
* **Web search + content** → `firecrawl_search`. Replaces Google search + individual page scraping in one call.
* **Site discovery** → `firecrawl_map` first to enumerate URLs, then `firecrawl_batch_scrape` the ones you need.
* **Full site crawl** → `firecrawl_crawl_params_preview` to estimate cost, then `firecrawl_crawl`.
* **Known URL list** → `firecrawl_batch_scrape`. More efficient than individual scrapes.
* **Structured extraction** → `firecrawl_extract` with a JSON schema or natural language prompt.
* **Complex web tasks** → `firecrawl_agent` with a natural language instruction. The agent navigates pages autonomously.

### Budget awareness

* Usage depends on the action. Scrape, crawl, batch scrape, and map bill per page. Search bills per result. Agent and extract use dynamic pricing.
* Scrape options increase per-page cost. PDF parsing, JSON format, Enhanced Mode, and Zero Data Retention each add to the base page cost, and they stack.
* Deepline credit pricing for these actions is generated from the provider pricing metadata and rendered on the public provider pages.
* Firecrawl can charge when infrastructure processes a request even if the target returns 403/404. Avoid blind retries of blocked URLs. inspect `metadata.statusCode`.
* Crawl defaults to `limit: 10000` and Firecrawl preflights available balance against that limit. Always pass an explicit lower `limit` unless a 10000-page crawl is intentional.
* `crawl_params_preview` is free and shows estimated usage before committing.

### Async operations

* `crawl`, `batch_scrape`, `agent`, and `extract` are async. They return a job ID immediately.
* The action handler polls automatically for up to 5 minutes and returns results when ready.
* For non-blocking usage, use the corresponding status-check action (`get_crawl_status`, etc.) to poll manually.
* Cancel long-running jobs with the cancel actions when results are no longer needed.

### Format recommendations

* Use `markdown` format for LLM consumption (default).
* Use `html` when you need the raw DOM structure.
* Use `links` to extract all hyperlinks from a page.
* Use `screenshot` when visual layout matters.

### Rate limits

* Standard rate limit is 15 requests/second.
* Batch and crawl operations are rate-limited server-side. The API handles queuing.

## Find an applicable action

Confirm your effective workspace, then search the live catalog:

```bash theme={"languages":{"custom":["/languages/deepline-prompt.json"]}}
deepline auth status --json
deepline search "Firecrawl" --type tools --json
```

Search returns candidates. Check the returned provider and operation rather than selecting a similar title automatically.

## Inspect the selected contract

Replace `<returned-tool-id>` with the identifier returned by discovery:

```bash theme={"languages":{"custom":["/languages/deepline-prompt.json"]}}
deepline describe <returned-tool-id> --json
```

These commands do not execute the selected provider action. Read its required inputs, supported access mode, pricing unit, outputs, and limits.

A connection does not establish that every action supports BYOK. Follow the provider-specific requirements above and the live action contract.

## Prepare one reviewed input

Use one authorized record that matches the inspected schema. Preserve its source identity and define the result you will accept.

Open the action reference for its payload and response shape. Provider-backed execution can incur usage or external effects.

For reusable logic, follow [Write Plays](https://deepline.com/docs/sdk-v2/plays.md). For direct-call recovery, use [Tool execution recovery](https://deepline.com/docs/sdk-v2/tool-execution-recovery.md).

> [!TIP]
>
> Continue only when the workspace, provider, required fields, acceptance rule, and applicable price are understood. Keep unavailable access separate from no-result evidence.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.