Irmu vs Firecrawl
Agent-ready markdown and crawling versus a full fetch, browser, screenshot and proxy stack.
Firecrawl is the clearest expression of the agent era of web data. You hand it a URL, it hands back clean markdown ready to drop into a prompt, and its /crawl, /map, /extract, /search and /agent endpoints cover most of what a retrieval pipeline needs without you writing a parser. The LangChain and LlamaIndex integrations are first-party, which matters when your ingestion layer already speaks those abstractions.
Irmu overlaps at the output layer — markdown, structured JSON and AI extraction are all there — but the product underneath is different. Irmu also ships a scriptable browser endpoint, screenshots, and an HTTP proxy port for code you do not want to rewrite. If your workload is retrieval over cooperative sites, the extra surface is dead weight. If it involves logged-in flows, hostile anti-bot, or an existing Playwright script, that surface is the whole point.
The honest split: Firecrawl is a context API, Irmu is an access layer with a context API attached.
What each one costs
Public list pricing, checked on 29 August 2026. Vendors change plans often — verify on their pricing page before you commit.
Irmu
$40 / month — 200,000 credits
Credits, one ledger for every API. Plain fetch is 1 credit, JavaScript rendering 5, premium residential 10, premium plus rendering 25, an AI extraction query 5, and geotargeting by country, region or city is free. A response from the target — including 404, 500 or an anti-bot block — is a successful request and is billed. Requests Irmu cannot serve for its own internal reasons, or to domains outside coverage such as .gov, are not billed.
Firecrawl
$19 / month (Hobby) — 5,000 credits
Credit-based, with the cost per call depending on the endpoint and the options you pass. Firecrawl's docs describe the model but do not publish a single flat multiplier table, so budget against your own traffic rather than a headline rate.
- Firecrawl's free plan is 1,000 credits a month with 2 concurrent requests and no card required; Irmu's free plan is 200 credits a month with 1 concurrent request, also without a card.
- Compare on credits per successful page for your own mix, not on plan price — a rendered, residential-proxied fetch and a plain markdown scrape are not the same unit of work anywhere.
- Firecrawl's failed-request billing policy was not published on the pages we checked. Irmu's is explicit: only Irmu-side failures and uncovered domains go unbilled.
Side by side
| Capability | Irmu | Firecrawl |
|---|---|---|
| Markdown output for LLMs | Yes, on every fetch | Yes, the core output format |
| Site crawl and sitemap discovery | Batch endpoint with webhook delivery | Dedicated /crawl and /map endpoints |
| Scriptable remote browser | Browser endpoint for clicks, scrolls and waits | Not published as a first-party endpoint |
| Screenshots | Full-page and viewport screenshot API | Available via scrape options |
| HTTP proxy port | proxy.irmu.com:8080 for existing code | Not offered |
| Residential IPs | Premium tier, 10 credits per request | Handled internally, not exposed as a tier |
| Geotargeting | Country, region and city at no extra cost | Location option on scrape |
| Agent framework integrations | llms.txt, markdown mirrors, agent rules files | First-party LangChain and LlamaIndex packages |
Which one fits your job
Choose Firecrawl when
- Your pipeline is retrieval over mostly cooperative documentation, blogs and marketing sites.
- You want LangChain or LlamaIndex to do the plumbing with a package that already exists.
- Whole-site crawling with automatic link discovery is the primary job, not one-off fetches.
- You are prototyping and the 1,000-credit free plan covers your evaluation.
Choose Irmu when
- Some targets need a real browser you can drive — click, scroll, wait for a selector.
- You already have Playwright, Scrapy or curl-based code and want a proxy port rather than a rewrite.
- You need residential IPs and city-level geotargeting as explicit, priced controls.
- You want one key and one credit ledger across fetch, browser, screenshot, extract and proxy.
What Firecrawl does well
- Clean markdown built for LLM pipelines
- First-party LangChain and LlamaIndex packages
- Crawl, map and agent endpoints out of the box
- Generous free tier for prototyping
Firecrawl is the better tool when the job ends at clean text for a model. Irmu is the better tool when getting to the page is the hard part and clean text is the last step.
The same request, both ways
Firecrawl returns markdown from a POST body; Irmu takes query parameters and returns HTML, markdown or JSON depending on the format you ask for.
Firecrawl
import requests
r = requests.post(
"https://api.firecrawl.dev/v1/scrape",
headers={"Authorization": f"Bearer {FIRECRAWL_KEY}"},
json={"url": "https://example.com/product/42", "formats": ["markdown"]},
timeout=90,
)
markdown = r.json()["data"]["markdown"]Irmu
import requests
r = requests.get(
"https://app.irmu.com/api/scrape",
params={
"api_key": IRMU_KEY,
"url": "https://example.com/product/42",
"format": "markdown",
"render_js": "true",
},
timeout=90,
)
markdown = r.textSwitching questions
Other comparisons
Irmu vs ScrapingBee
A closely matched credit model — the difference is browser control and the proxy port.
Irmu vs ScrapingDog
Similar entry pricing — the split is dedicated site scrapers versus general-purpose access.
Irmu vs ScraperAPI
Structured e-commerce endpoints and no-code pipelines versus a general access layer.
Irmu vs ZenRows
Two similar credit-based platforms — the difference is agent tooling versus browser and proxy depth.
Irmu vs Bright Data
The largest proxy network in the market versus a simpler, credit-priced access layer.
Irmu vs Apify
A marketplace of pre-built Actors versus a direct API you call from your own code.
Irmu vs Oxylabs
Per-target enterprise pricing and strict pay-for-success versus a single credit ledger.
Try Irmu alongside Firecrawl
200 free credits every month, no card required — benchmark on your own targets.