Add AG2 (formerly AutoGen) web research pipeline example

[VasiliyRad]

Adds ag2-web-research/ — a standalone multi-agent web research pipeline using AG2 v0.11+ GroupChat with Firecrawl search, scraping, and deep research tools. Three specialist agents (searcher, scraper, reporter) run in round-robin order with @register_for_llm / @register_for_execution tool registration.

Why: Firecrawl's examples repo shows how to integrate with Google ADK and CrewAI, but has no AG2 (formerly AutoGen) example. AG2 is the community-maintained successor to Microsoft AutoGen with a large developer audience that overlaps with Firecrawl's user base. The @register_for_llm / @register_for_execution split — separating tool description from tool execution — is architecturally distinct from both ADK (tools=[]) and CrewAI (BaseTool subclass), and maps cleanly to Firecrawl's API surface. No changes to Firecrawl's API wrappers or existing examples are needed.

Files added: ag2-web-research/ (4 files — main.py, requirements.txt, .env.example, README.md)

[vercel]

@VasiliyRad is attempting to deploy a commit to the Firecrawl Team on Vercel.

A member of the Team first needs to authorize it.

[VasiliyRad]

@ericciarla, could you please review this PR? It adds an AG2 multi-agent web research pipeline that uses Firecrawl's search and scrape APIs through a three-agent GroupChat (searcher, scraper, reporter). AG2 is the community-maintained successor to Microsoft's AutoGen framework — actively developed with native async, tool registration, and multi-agent orchestration.