Web-scraping agent Firecrawl serves 5 billion requests for AI workflows; discloses no governance, audit trail or insurance terms

Firecrawl, a web-scraping and search API marketed to AI agents, says it has "served more than 5 billion requests powering deep research agents, RAG pipelines, lead enrichment, and AI workflows across the live web," according to its website. The company reports more than 1.25 million developers and 150,000 companies building on the product, and says it delivers "P95 latency of 3.4s across millions of searches and scrapes, built for real-time agents and dynamic apps."

The product ships an official Model Context Protocol server, letting agents inside Cursor, Claude, Windsurf and other MCP-compatible tools "search, scrape, and interact with the web directly," the company says.

An AI Dispatch review of Firecrawl's public disclosures found no evidence on any of the five dimensions of the trust-gap rubric this outlet applies to agentic AI and AI-supplier products: runtime governance, audit trail, compliance, outcome-based settlement or insurance and indemnity. No scope, spend or authority limit on the agent is published. No logging, record-keeping or reviewability claim appears on the site. No certification, framework mapping or auditor report is disclosed. No cover, named carrier, policy limit or customer indemnity is published.

On billing, Firecrawl says "we currently do not offer a pay-per-use plan, instead you can upgrade to our Hobby, Standard or Growth plans for more credits and higher rate limits" — payment by subscription tier regardless of outcome, not settlement tied to a verified result. The company's stated recourse for a billing dispute is customer support: "If a charge looks incorrect, reach out to help@firecrawl.com and we'll take a look."

Firecrawl also describes itself as open source at scale, saying it is "the largest open source repo in the space with over 150K GitHub stars, making it one of the top 100 repos on GitHub."

The absence of governance and insurance disclosures does not necessarily reflect what Firecrawl has built internally — only what it has published. AI Dispatch scores public disclosure, not undisclosed practice, and treats "no public information" as a citable finding rather than a gap to be assumed away.