Extract Web Data at Scale With Parallel Agents

It's been an exciting week at Firecrawl. Following the release of our Firecrawl Skill and CLI, we're now bringing parallel processing to /agent, letting you batch hundreds or even thousands of queries simultaneously.

• Parallel Agents for Firecrawl /agent endpoint - Batch thousands of /agent queries simultaneously to enrich companies, research competitors, or build product datasets at scale
• Intelligent waterfall - Parallel Agents tries instant retrieval first with Spark-1 Fast, then automatically upgrades to Spark-1 Mini whenever full agent research is needed
• Real-time spreadsheet interface - Work in familiar CSV format instead of JSON only. Watch cells populate with instant visual feedback
• Zero configuration - Input your data schema, write one prompt, hit run. No workflow building required

With Parallel Agents, powered by our newest model Spark-1 Fast, you can batch enrich massive datasets in minutes. Just write your prompt, hit run, and watch as data populates in real-time in our playground.

We built Firecrawl /agent, and developers loved it. /agent searches, navigates, and gathers data from even the most complex websites, finding information in hard-to-reach places and discovering data anywhere on the internet.

But /agent processes one query at a time.

Need to enrich 500 companies with funding stages and contact information? That's 500 sequential queries. Compare features and capabilities across 200 competitors? Same problem.

You could build workflows in tools like Clay, connecting multiple sources and configuring fallback logic. But that takes significant time and effort.

With /agent's parallel processing, you can now process thousands of queries simultaneously. No setup required.

/agent tries the fastest method first for every cell, then auto-upgrades to full research only when necessary.

Fast path - Instant retrieval for common queries, now powered by our new Spark-1 Fast model. Think known company information, funding stages, standard contact details.

/agent path - Full Spark One Mini research for complex queries. This includes hard-to-find information, dynamic content, and multi-step navigation.

You don't need to configure this. Firecrawl handles everything automatically.

Work in CSV (spreadsheet) or JSON formats. See data populate as queries complete.

Cells fill in real-time.

• Green = success.
• Yellow = in-progress.
• Red = failure.

Write one prompt. Hit run. Firecrawl handles routing, failure handling, and method selection automatically.

Firmographic enrichment

Input company names, get funding stages, employee counts, and contact information populated automatically. Fast retrieval handles known companies instantly. /agent finds information for startups that aren't in standard databases.

Competitive research

Compare features, capabilities, and market positioning across competitors to build intelligence reports. /agent navigates sites, clicks through pages, and extracts data to structure competitive comparisons.

Product data extraction

Extract entire catalogs with specs, pricing, and reviews from e-commerce brands at scale. /agent handles pagination, clicks into product pages, and structures thousands of products into clean datasets.

Start enriching at scale with predictable pricing:

• Fast path - 10 credits per cell for Spark-1 Fast instant retrieval
• Agent path - Standard Spark-1 Mini /agent pricing for full research

The intelligent waterfall optimizes costs automatically. You only pay for full Spark-1 Mini agent research when Spark-1 Fast retrieval fails.

1. Head to the Agent Playground
2. Upload a .csv or ask /agent to enrich multiple entities (e.g., "Find the funding stage, employee count, and contact email for these 50 companies")
3. The playground automatically determines when your use case is a good fit for Parallel Agents
4. Share your projects in the Firecrawl Community

What is parallel processing for Firecrawl Agent?

The /agent endpoint now has batch processing capabilities that let you run hundreds or thousands of web data queries simultaneously, viewable in CSV or JSON format.

How does the intelligent waterfall work?

Each cell tries instant retrieval first with Spark-1 Fast. If that fails, it automatically upgrades to Spark-1 Mini for full /agent research, so you only pay for deeper work when needed.

What inputs can I use?

You can paste a CSV-style spreadsheet with column headers or provide JSON data plus a prompt that describes the fields you want filled in.

How are credits billed?

Fast path requests use 10 credits per cell for Spark-1 Fast. When a query upgrades to full /agent research, it uses standard Spark-1 Mini /agent pricing.

Where can I access Agent?

You can try /agent in the Agent Playground.