Firecrawl - Extract data action

The Extract data action runs an Agent extraction job in Firecrawl. A prompt is required, and you can optionally add a JSON schema to further constrain the extracted data. URLs are optional. Agent searches and navigates the web on its own if you leave URLs blank.

WAIT FOR COMPLETION

This action submits the job by default, then polls every 60 seconds until the job's status is completed or failed. Select No for Wait for completion to return the job ID immediately instead, then use the Get extract job status action to check on it later.

Input

The following input fields are the most commonly used. Refer to the Firecrawl Agent API documentation for the complete list of input fields.

Input fieldDescription
PromptEnter a natural-language description of the data to extract. The maximum is 10,000 characters.
JSON schemaOptional. Enter a JSON schema describing the data to extract. This schema is sent to Firecrawl only. Add matching fields in Custom output fields to get datapills for the extracted data.
URLsOptional. Enter the URLs to focus the Agent on.
Enable web searchOptional. Select Yes or No. Select No to keep the Agent only on the URLs you provide. The default is Yes.
Max creditsOptional. Enter a credit cap for this run. The job fails with no data if the cap is reached. This connector default is 100. The Firecrawl API default is 2500.
ModelOptional. Select the model to use. Options are Spark 2, Spark 1 Mini, or Spark 1 Pro. The default is Spark 2. Firecrawl accepts the Spark 1 options but routes crawl requests to Spark 2.
EffortOptional. Select the reasoning budget for Spark 2. Options are Low, Medium, or High.
Wait for completionOptional. Select Yes or No to poll until the job finishes and returns the result. The default is Yes.
Max poll minutesOptional. Enter the maximum minutes to poll. Valid values are 1-60. The default is 10. This field is only available when Wait for completion is Yes.
Custom output fieldsOptional. Add fields to create datapills for extracted data fields that the default output doesn't include.

Output

Output fieldDescription
SuccessA true or false boolean response that indicates if the request was successful.
StatusThe job's status. Options are processing, completed, or failed. Canceled jobs also report failed.
ModelThe model used for the job.
EffortThe reasoning budget used for the job.
ModeThe Agent mode used for the job, for example extract.
Expires atThe date and time the job's results expire.
ErrorThe error message, if the job failed.
MessageA message describing the result of the job.
DataThe extracted data, matching your prompt or schema.
Created atThe date and time the job was created.

Last updated: