Artificial intelligence news.
Straight from the original announcements.

Firecrawl MCP v3 Gives Agents Access to 3,000+ Tools

November 12, 2026By the AI News Bites Editorial TeamSource: Firecrawl
Firecrawl MCP v3 Gives Agents Access to 3,000+ Tools
Image: Firecrawl

Firecrawl has released MCP v3, a connection that gives agents access to more than 3,000 tools and can continue searching when one provider cannot find an answer. The update also raises rate limits, increases MCP usage and lets agents delegate work to parallel subtasks.

Key points
  • MCP v3 provides access to 3,000+ Firecrawl tools through one connection.
  • A provider waterfall continues searches after a first “not found” result.
  • Rate limits are up to 60x higher and MCP usage is 4x greater.
  • Agents can delegate work to subtasks that run in parallel.
  • The update works in Claude, ChatGPT and any harness that speaks MCP.

One connection to Firecrawl’s full catalog

Firecrawl MCP v3 connects an agent to the full Firecrawl Alexandria catalog, including company data, people data, research and government records. Search, Scrape and Parse are also available through the connection. Instead of wiring providers one at a time, an agent describes what it needs and the MCP finds the appropriate tool.

According to Firecrawl, provider discovery is about 8x faster than in v2. The company says this allows an agent to spend more time on its task rather than searching for the tool needed to complete it.

Firecrawl says MCP v3 is live today in every harness at alexandria.firecrawl.dev/go/mcp. It works in Claude, ChatGPT and any harness that speaks MCP.

Provider fallback aims to reduce dead ends

A central change in v3 is a people enrichment waterfall. When the first provider returns “not found,” the system continues to the next source until it finds the requested record. The agent makes one call and receives one answer, rather than having to ask several providers separately.

Firecrawl presents work email lookup as one example. The company says the information may exist with another provider even when the first source returns no result. The waterfall is designed to keep the lookup moving instead of ending at the first unsuccessful search.

The update also targets larger agent workloads. Search through the MCP is about 2x faster, while rate limits are up to 60x higher. Firecrawl says users get 4x more usage through the MCP, and that every plan includes free MCP usage that refreshes daily.

Agents can split work into parallel subtasks

MCP v3 allows an agent to spawn subtasks that run in the background while the main task continues. For example, an agent researching ten prospects can split the work into ten subtasks, run them in parallel and use the results as they arrive.

Firecrawl says this additional capacity is intended for agents that make many calls instead of working one call at a time. The company gives a research agent checking 200 companies in an afternoon as an example of a workload that could benefit from the higher limits.

Users can connect Firecrawl to ChatGPT, Claude, Cursor and more in one click. Firecrawl directs users to the MCP connection page to try the service and asks them to share feedback on X at @firecrawl.

Related news

Products and tools
Products and toolsNovember 14, 2026, 02:41

GreenCore Solutions changes name to WeFetch AI Corp.

GreenCore Solutions Corp. has completed its change of name to WeFetch AI Corp., as the company positions its barcode technology for AI-driven shopping. Its lead product…

Products and tools
Products and toolsNovember 12, 2026, 19:21

RemotePC Adds MCP Integration for AI-Assisted Computer Management

RemotePC has introduced an integration that lets customers use ChatGPT, Claude and other compatible AI assistants to monitor remote computers and perform supported…

Firecrawl introduces Alexandria for answers from official sources
Products and toolsNovember 11, 2026

Firecrawl introduces Alexandria for answers from official sources

Firecrawl has introduced Alexandria, a system that helps agents retrieve information from official sources before falling back to web scraping. It is already available…