Firecrawl Parse

Usage-based pricing
ae_firecrawl_parse

Parse one local document into Markdown, HTML, raw HTML, links, a summary, JSON or query answers using Firecrawl. Supported file extensions are .html, .htm, .xhtml, .pdf, .docx, .doc, .odt, .rtf, .xlsx and .xls. Input and workflow • Provide exactly one nonempty string: filePath or uploadRef. In local MCP mode, filePath points to a file on the MCP server filesystem. In hosted mode, filePath refers to the caller's file but is not read or checked by the hosted server: the first call returns upload instructions. After the caller completes the upload, call again with the returned short-lived uploadRef and the desired parsing options. • Hosted mode is a two-call upload handoff, not an asynchronous parse job with status polling. This tool exposes no separate job-status/result retrieval lifecycle. Upload instructions alone are not parsed content. Options • formats: an array of markdown, html, rawHtml, links, summary, json and/or query. For structured extraction, jsonOptions accepts prompt and schema (a JSON schema object). For query answers, queryOptions requires prompt (up to 10000 characters) and mode: directQuote or freeform. Although the schema lists freeform as the default, mode is required when supplying queryOptions. • For PDFs, include parsers:["pdf"]. pdfOptions.maxPages optionally limits parsing to an integer from 1 to 10000 pages. • Hosted phase-one contentType overrides the MIME type; otherwise it is inferred from the file extension without reading the file. declaredSizeBytes is optional and must be an already-known integer size from 1 to 9007199254740991 bytes; this range is not a promise that uploads of that size are accepted. • includeTags and excludeTags select HTML tags; onlyMainContent requests main content only; removeBase64Images removes base64-encoded images from output. • redactPII requests redaction of personal information. zeroDataRetention requests no retention after processing, requires an eligible authenticated account, and should be omitted for anonymous keyless use. • proxy accepts basic or auto only. skipTlsVerification requests skipping TLS certificate verification when retrieving file resources, reducing transport verification. storeInCache is an exposed cache-storage request, not a documented cache guarantee. maxAge is ignored: parse never reuses or stores indexed content. All options other than the filePath/uploadRef choice are optional; extra unsupported properties are not accepted. Returns The response is an object whose fields vary by phase and outcome. Hosted phase one can return mode, upload instructions (uploadUrl, method, headers, fields, command, expiresAt, maxSizeBytes and uploadRef), nextToolCall as {name, arguments}, notes and message. Final data contains parsed content in the requested formats and may include data.metadata.scrapeId. Other possible fields are success, error, warning, raw for a non-JSON response body, agent_hints and conversationId. Fields are not guaranteed to be present, and no fixed mode/status values or fixed inner data layout are specified. Follow the returned upload expiry and maximum size rather than assuming global limits. Limits One file per parse; repeat calls for multiple files. Remote web URLs belong in the separate firecrawl_scrape operation. Interactive actions, screenshots, branding and change tracking are not supported; waitFor, location and mobile are not exposed by this interface. Do not send filePath together with uploadRef. For specific data points, JSON with a schema is appropriate; for complete document text, Markdown is appropriate.

Productivity
Usage-based pricing

Pricing

The charge per call is (max(1.2 × document parsing service usage units, 0)) AE credits.

Calls using a local file path are free. Hosted parsing calls after upload are charged according to actual processing usage.

Parameters

Optional parameters use the platform defaults when omitted.

ParameterTypeRequiredDescriptionDefault
proxyenum (basic | auto)NoNo parameter description.—
maxAgenumberNoIgnored: parse never reuses or stores indexed content.—
formatsarrayNoOutput formats to generate from the parsed file, such as markdown, HTML, JSON, summary, links, or query results.—
parsersarrayNoParsers to apply to the file. Include "pdf" when parsing PDF files.—
filePathstringConditionalPhase 1 only: path to the local file on the caller/harness machine. Hosted MCP will not read or stat this path; it is used only to produce upload instructions.—
redactPIIbooleanNoRedact personally identifiable information from returned content.—
uploadRefstringConditionalPhase 2 only: short-lived upload reference returned by phase 1 after the local PUT upload completes.—
pdfOptionsobjectNoOptions for the PDF parser.—
pdfOptions.maxPagesintegerNoMaximum number of PDF pages to parse.—
contentTypestringNoPhase 1 MIME type override. If omitted, the server infers it from the file extension without reading the file.—
excludeTagsarrayNoHTML tags to exclude from the parsed output.—
includeTagsarrayNoHTML tags to include in the parsed output.—
jsonOptionsobjectNoOptions for extracting structured JSON from the file.—
jsonOptions.promptstringNoInstructions describing the data to extract as JSON.—
jsonOptions.schemaobjectNoJSON schema that defines the structure of the extracted data.—
queryOptionsobjectNoOptions for querying the parsed file.—
queryOptions.modeenum (directQuote | freeform)YesNo parameter description.freeform
queryOptions.promptstringYesQuestion or query to run against the file content.—
storeInCachebooleanNoStore the parsing result in the cache.—
onlyMainContentbooleanNoReturn only the main document content.—
declaredSizeBytesintegerNoOptional phase 1 size declaration. Hosted MCP does not stat the file; provide this only if the caller already knows it.—
zeroDataRetentionbooleanNoDo not retain file data after processing.—
removeBase64ImagesbooleanNoRemove base64-encoded images from the parsed output.—
skipTlsVerificationbooleanNoSkip TLS certificate verification when retrieving file resources.—

Invocation example

Copy the request and explore its response example.

POST/agent-api/v1/tool/executeX-Api-Key
Request coderequest.sh
curl --request POST \
  --url 'https://agentearth.ai/agent-api/v1/tool/execute?tool_name=ae_firecrawl_parse' \
  --header "X-Api-Key: $AGENTEARTH_API_KEY" \
  --header 'Content-Type: application/json' \
  --data-raw '{
  "params": {
    "formats": [
      "markdown"
    ],
    "onlyMainContent": false,
    "parsers": [
      "pdf"
    ],
    "removeBase64Images": true,
    "uploadRef": "eyJ2IjoxLCJkcml2ZXIiOiJnY3MiLCJ1cGxvYWRJZCI6IjAxYTFhNDc4LTJjZmQtNzA0Mi1iMzk3LTg5ZmQ4ODRlYzgyNyIsInRlYW1JZCI6IjFlZWRlZWMxLTA5NjEtNDAyMy1hMTdiLTYyOTk5MGM5ZjgzMCIsIm9iamVjdFBhdGgiOiJwYXJzZS11cGxvYWRzLzFlZWRlZWMxLTA5NjEtNDAyMy1hMTdiLTYyOTk5MGM5ZjgzMC8wMWExYTQ3OC0yY2ZkLTcwNDItYjM5Ny04OWZkODg0ZWM4MjcvMS0yNjBINjE1Mjk0S0NfXzFfLnBkZiIsImZpbGVuYW1lIjoiMS0yNjBINjE1Mjk0S0NfXzFfLnBkZiIsImNvbnRlbnRUeXBlIjoiYXBwbGljYXRpb24vcGRmIiwiZXhwaXJlc0F0IjoxNzkzNzYxMzAxNjkzLCJtYXhCeXRlcyI6NTI0Mjg4MDB9.o2mmkgjqWlnCTrw9f3IDriG073ZmH3VmGniVcBaumFo"
  }
}'
Response exampleresponse.json
JSONPartial content
{
  "conversationId": "3c2835cc-4b10-44de-8581-0a8c25bf52ec",
  "data": {
    "markdown": "## Horizontal Laminar Flow Hood\n\n- Adopting ultra-thin partitionless filter to minimize the size of static pressure box, together with stainless steel countertop and glass baffle, the whole working area looks spacious and bright.\n\n- Conveniently removable ",
    "metadata": {
      "concurrencyLimited": false,
      "contentType": "application/pdf",
      "numPages": 1,
      "proxyUsed": "basic",
      "url": "https://parse.firecrawl.dev/uploads/1-260H615294KC__1_.pdf"
    }
  },
  "success": true
}

This example shows part of the response. Copied content is valid JSON.