OpenRouter adds Firecrawl web search with full-page markdown grounding

• OpenRouter added Firecrawl as a web search provider, and firecrawl's launch post says the integration grounds model responses in full scraped pages rather than snippet-only search results.
• The setup path is not a new product surface. firecrawl's settings reply points users to OpenRouter's existing Plugin settings flow.
• firecrawl's demo post frames the output format as clean markdown, which is the useful bit for engineers who want models reasoning over page content instead of search summaries.
• The launch also piggybacks on Firecrawl's acquisition funnel, because firecrawl's announcement says new sign-ups get a free Hobby plan with 100,000 credits.

You can jump straight to OpenRouter's plugin settings page, and firecrawl's launch video shows the integration as a simple provider toggle rather than a separate agent product. Wes Roth's earlier web-agent writeup is a useful contrast point: Firecrawl is also building a broader search-scrape-interact agent framework, but this OpenRouter ship is the narrower retrieval layer.

The main change is the retrieval format. According to firecrawl's launch post, Firecrawl searches the live web, scrapes the selected pages, and returns them as clean markdown for any OpenRouter model using the provider.

That is a different primitive from the usual snippet-first search add-on. The tweet's phrasing, "full web pages instead of search snippets," is the whole story here, because it moves the grounding context from short result summaries to page-level content.

The integration lives inside OpenRouter's existing plugin UI. In firecrawl's follow-up reply, the setup instruction is simply to enable Firecrawl in OpenRouter's Plugin → Web Search settings.

That makes this look more like a new backend for an existing search surface than a standalone product launch. OpenRouter's retweet amplified the same announcement without adding extra workflow or pricing detail.

Wes Roth's post described Firecrawl's separate open-source "web-agent" framework a day earlier as a model-agnostic system for autonomous search, scrape, and interact loops, with a scaffold command, Plan-Act execution, and parallel sub-agents. That broader stack gives some context for why Firecrawl is a plausible search backend here.

The OpenRouter integration is much smaller in scope. It exposes Firecrawl as a search provider, not the full agent framework Roth described.

The one concrete adoption incentive in the launch post is on Firecrawl's side, not OpenRouter's. firecrawl's announcement says new sign-ups get a free Hobby plan with 100,000 credits.

That free-tier detail did not appear in OpenRouter's retweet, which mostly signals distribution. The promotional hook came from Firecrawl itself.