Back to blog
Product·August 24, 2026·5 min read

Zero Touch: A Web Scraping API With No API Key

We removed the API key from our free tier. One curl gets you scraped pages, live site actions, and an MCP server with no signup and no OAuth, metered per IP with the upgrade path built into the response.

A

Arun Singh

Anakin Team

Zero Touch: the API key is a human ritual. No key, no signup, one call.

The fastest way to evaluate an API is to call it. Every step between a developer and that first response exists for the vendor's convenience, not the developer's. The signup form, the email verification, the dashboard, the key: each one is friction we invented. And for the newest consumers of web data it's worse than friction, because your AI agent doesn't have an email address. It can't fill in a signup form, wait for a verification link, or click through an OAuth consent screen.

So we removed all of it.

Zero Touch is Anakin's no-key tier: scrape any URL, run read-only actions on more than 970 sites, or connect over MCP, all with no API key, no signup, and no card. Your first request is the integration test:

curl -s https://api.anakin.io/v1/url-scraper/scrape \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com"}'

That is the entire onboarding.

What you get without a key

Zero Touch isn't a demo sandbox. It serves the same production endpoints as the paid API, minus the auth header:

Scrape any URL. `POST /v1/url-scraper/scrape` returns the page inline as clean markdown, with the raw and cleaned HTML in the same JSON response and an optional headless browser (useBrowser) for JavaScript-heavy pages. It runs behind the same anti-bot handling, proxy selection, and rendering pipeline our paid traffic uses.

Run live site actions. POST /v1/wire-run executes read-only actions from our Wire catalog, which covered 973 sites and 5,279 actions on 25 Oct 2026 and keeps growing. GET /v1/wire/catalog and GET /v1/wire/resolve are open too, so an agent can discover what's runnable before it runs anything. On 25 Oct 2026, resolving "hacker news stories" pointed to the hn_stories action, and one keyless call returned the top stories inline, no polling:

curl -s https://api.anakin.io/v1/wire-run \
  -H "Content-Type: application/json" \
  -d '{"action_id": "hn_stories", "params": {"limit": 3}}'

# response, trimmed (run on 25 Oct 2026)
{
  "status": "completed",
  "data": {
    "type": "top",
    "stories": [
      {"title": "A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis",
       "score": 31, "comment_count": 2,
       "hn_url": "https://news.ycombinator.com/item?id=50233779"},
      ...
    ]
  },
  "credits_used": 2,
  "trial": {"remaining_credits": 288, ...}
}

Connect over MCP, without OAuth. Point any MCP-capable agent (Claude, Cursor, or your own) at mcp.anakin.io/mcp and it gets the keyless tool surface immediately: scrape, wire_discover, wire_catalog and wire_read_action. No OAuth dance, no token exchange, no consent screen. When you do have a key, add it as a bearer token and the same server upgrades in place to the full toolset.

The free tier is an HTTP header

Most free tiers live in a dashboard you have to log into. Ours lives in the response. These are the headers from a keyless scrape of example.com we ran on 25 Oct 2026:

HTTP/2 200
x-trial-credits-remaining: 291
x-anakin-signup: https://anakin.io/signup?source=keyless
x-ratelimit-limit: 20
x-ratelimit-remaining: 19
x-ratelimit-reset: 60

Every keyless call is metered per IP: 300 credits a month, with burst and concurrency caps to keep the tier honest. The x-ratelimit-* headers above are the rate cap reporting on every call, and that scrape cost 1 credit. The body carries the same signal in machine-readable form (trimmed):

{
  "status": "completed",
  "markdown": "This domain is for use in documentation examples without needing permission. ...",
  "trial": {
    "remaining_credits": 291,
    "signup_url": "https://anakin.io/signup?source=keyless"
  }
}

This matters more for agents than for people. A human checks a dashboard. An agent budgets from the header. When the credits run out you don't get a silent failure or a dropped connection. You get a 402 with the signup URL in the payload. The graceful exit is part of the API contract.

And when you outgrow the tier, there's no migration: add an X-API-Key header to the same endpoints and everything upgrades in place. Nothing you built against the keyless tier gets thrown away. A free account also opens what the keyless tier leaves out: 300 signup credits, write actions, crawl, agentic search, monitors, and account-connected Wire actions.

Why we think keys are the wrong default

API keys solve real problems: attribution, billing, abuse control. But for a first request they solve our problems, not yours. Per-IP metering with tight burst caps handles the abuse case for a trial tier, and the response headers handle attribution back to the caller. What is left is pure onboarding friction, and it lands hardest on the two audiences we care most about:

  1. Developers evaluating tools. If it takes a signup to find out whether the scraper handles your target site, most people never find out. A curl that either works or doesn't is a more honest pitch than any landing page.
  2. AI agents evaluating tools at runtime. Agents increasingly discover and select APIs on their own. They read llms.txt, scan catalogs, and try endpoints. An agent that hits a signup wall moves on to the next tool in its list. We expect keyless tiers to become table stakes for agent-facing infrastructure, and we think that's the right direction.

That's also why the whole tier is spelled out in our llms.txt in plain language: the no-key endpoints, how metering works, and the MCP server. If your agent reads one file about us, that one is enough to start working.

The boring parts, done properly

A no-key tier is easy to launch and easy to get wrong. Here's where the engineering went:

  • Per-IP monthly metering with burst and concurrency limits: generous enough to evaluate honestly, tight enough that the tier cannot be farmed.
  • A global monthly budget on the whole tier, controlled by us, so keyless traffic can never affect paid customers' capacity.
  • Fail-safe design. The metering layer degrades gracefully rather than taking your request down with it.
  • The same pipeline as paid. There is no cut-down trial engine. What you evaluate is what you would buy.

FAQ

Is there a web scraping API that works without an API key?

Yes. Anakin's Zero Touch tier lets you call POST https://api.anakin.io/v1/url-scraper/scrape with no API key, no signup, and no card. You get 300 free credits per IP each month, and every response reports your balance in its headers.

How do you use the free tier?

Send a request. There is no step two: no account, no form, no key. Each response includes x-trial-credits-remaining so you always know your balance.

What happens when free credits run out?

You get an HTTP 402 with a trial object containing a signup URL. Sign up for a key and the same endpoints keep working with higher limits, with no code changes.

Can AI agents use it without OAuth?

Yes. The MCP server at mcp.anakin.io/mcp exposes the keyless tool surface with no OAuth flow, and the REST endpoints work with no auth header. The tier is documented in anakin.io/llms.txt for agent discovery.

Is keyless traffic slower or lower quality than paid?

No. Keyless requests run through the same scraping pipeline, anti-bot handling, and proxy infrastructure as paid traffic. The difference is the metering, not the engine.

Try it now

The best way to judge it is the way this post started: call it. One request, from your terminal:

curl -s https://api.anakin.io/v1/url-scraper/scrape \
  -H "Content-Type: application/json" \
  -d '{"url": "https://news.ycombinator.com"}'

Full details on the Zero Touch product page and in the docs.