Summary: On August 20, 2026, Firecrawl became a first-class web data provider inside Eden AI, letting users combine live web ingestion with more than 500 models under one Eden AI key and bill. Seven Firecrawl capabilities went live, spanning scraping, search, crawl, batch scraping, structured extraction, and deep research. Direct use of Firecrawl’s raw capabilities still requires a separate Firecrawl key.
The Change: Firecrawl Now Lives Inside Eden AI
Anyone building RAG pipelines, agents, or research tools needs two ingredients: live data pulled off the web, and a model capable of reasoning over it. Traditionally that has meant two vendors, two dashboards, and glue code holding them together. Firecrawl handles the first half—it converts any website, JavaScript-rendered and dynamic pages included, into clean, LLM-ready data. Eden AI takes the second half, putting Claude, GPT, Gemini, Mistral, and 500+ other models behind one API and one bill.
Firecrawl announced on August 20, 2026 that it is now available directly inside Eden AI as a first-class web data provider. No separate sign-up, and no second API key to chase down. Firecrawl sits in that same catalog under the web feature, with seven capabilities live:
- Web Scraping (
web/scraping/firecrawl) – Fetch a single URL and receive clean markdown, text, or HTML. - Web Search (
web/search/firecrawl) – Search the web and get ranked results carrying title, URL, and content. - Site Map (
web/map/firecrawl) – Surface the list of URLs found on a site. - Web Crawl (
web/crawl_async/firecrawl) – Starting from a URL, crawl an entire site and retrieve the content of its pages. - Batch Web Scraping (
web/batch_scrape_async/firecrawl) – Scrape a long list of URLs within one request. - Structured Extraction (
web/structured_extraction_async/firecrawl) – Convert pages into JSON against a prompt, a schema, or the two together. - Deep Research (
web/research_async/firecrawl) – Carry out multi-step web research on a query and get findings back with their sources.
How Calls Work: Sync and Async
Each call reaches the standard Eden AI endpoint with one EDEN_AI_API_KEY, following the same feature/subfeature/provider pattern the rest of the platform uses. Scraping, search, and map are synchronous: you POST to /v3/universal-ai and the content arrives in the response:
curl -X POST https://api.edenai.run/v3/universal-ai \
-H "Authorization: Bearer $EDEN_AI_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "model": "web/scraping/firecrawl", "input": { "url": "https://firecrawl.dev" } }'Crawl, batch scrape, structured extraction, and deep research, by contrast, run as jobs. You POST to /v3/universal-ai/async, then poll /v3/universal-ai/async/{public_id} until the status flips to success. Responses arrive in the same envelope every other Eden AI feature uses, with a cost field attached to each call, so web data and model calls end up on one bill.
The entire catalog can be browsed before you sign up for anything, including every Firecrawl model string, its mode, and its pricing:
curl https://api.edenai.run/v3/info/webIn Practice: From RAG to Autonomous Research
Several everyday build scenarios get more direct thanks to this integration:
- RAG on live web data. Crawl or scrape a set of sources with Firecrawl, embed the markdown, then answer questions with any model in the catalog. A single key covers the whole loop.
- Web-augmented agents. Hand an agent
searchandscrapingas tools beside its LLM, so that when it lacks an answer it goes and finds one rather than guessing. - Structured extraction at scale. Batch scrape a list of URLs, run structured extraction with a prompt or schema, and let a model post-process the JSON. Comes in handy for lead enrichment, competitive tracking, and catalog monitoring.
- Market monitoring. Put a domain on a map-and-crawl schedule, then have a model summarize what changed since the previous run.
- Autonomous research. Give deep research a plain-language brief and it searches and reads the web on its own, returning what it found along with sources, shaped to a schema whenever you pass one.
Constraints and Trade-offs
One distinction is worth flagging: this is not the same as Eden AI’s LLM web search option. The web_search_options flag switches on a model’s built-in search while an LLM call is running. The Firecrawl integration, by contrast, is a separate web feature that hands back actual page content, leaving it up to you what gets scraped, what gets crawled, and what gets passed to the model.
If Eden AI is already how you manage model calls, adding Firecrawl beneath the same key saves a great deal of integration work. Should you need to call Firecrawl’s raw capabilities directly, though, a separate Firecrawl key is still required.
Eden AI is a French company, and its gateway operates on EU infrastructure, offering GDPR-compliant processing under a single contract that spans all providers. Every provider lists its region in the public info endpoint, so where a model runs can be checked before you call it. For European teams, that can trim vendor review overhead.
The Upshot: One Key Runs the Whole Pipeline
With the Firecrawl + Eden AI integration, a complete AI pipeline—web data ingestion plus model inference—can now be put together with one API key and one bill. For product builders, that removes a significant piece of glue code along with a chunk of vendor management. The public info endpoint lets pricing and regions be weighed before anything is committed. Already on Eden AI? This is a low-friction way to layer in live web data. Not on it yet? It is a compelling reason to think about consolidating your stack.