Microlink alternative

A Microlink alternative for agents and content pipelines

Microlink turns a URL into data for apps and agents, and does it well. The comparison worth making is what the same call returns, what it costs when the page fails, and how an agent is expected to hold it.

Stay where you are if

Your code depends on specific enrichment fields. We do not return them field for field, and checking that costs less than migrating.

Move here if

Text, metadata and pixels have to describe the same render, and your agent needs error codes with a documented retryable flag.

What moving looks like: Check the fields you consume against the options reference first. For an agent, the published skill carries the request shape and the retryable codes.

microlink alternative · request and resultPAGECAPTURE / 01
The same job, in our contract
curl -G https://api.pagecapture.dev/v1/take \
  -H "X-Access-Key: $PAGECAPTURE_KEY" \
  --data-urlencode "url=https://example.com" \
  -d "format=markdown" \
  -d "response_type=json" \
  -d "metadata_open_graph=true" \
  -o page.json
Capture result
succeeded
https://example.com
Noma
Shop now
NEW COLLECTION

Objects for slower mornings.

Thoughtful essentials, made in small batches from natural materials.

Explore the collection
No. 01 · Hinoki
PLANT-BASEDMADE IN SMALL BATCHESREFILLABLE
Artifact
PNG · 1440×900
Billing
1 credit
Engine
stable

Text and pixels from one render

Markdown, metadata and an image come from the same capture, so an enrichment pipeline is not comparing two different loads of the page.

Built for agents without a separate path

An MCP server, a published agent skill and llms.txt all describe the same HTTP API. The channel is recorded from a closed allow-list and never enters the cache key.

Errors an agent can act on

A stable error_code with a documented retryable flag, so an agent retries what is worth retrying instead of looping on a permanent failure.

Nothing billed for a page that did not load

Enrichment pipelines meet hostile pages constantly. A failed fetch produces no credit here.

What this costs here

Priced on successful captures that miss cache

Every number in this table is computed from the published plans — failures and organization-scoped cache hits are removed before the plan is chosen, because they are not billable.

ScenarioBillable successesSmallest included planPublished monthly price
5,000 requests per month4,410Growth$47.00

Assumes 10% cache hits and 2% failures. Change either assumption and the billable count changes with it — that is the point of charging for success.

What these vendors publish

Microlink, in their own numbers

These figures are quoted from each vendor's own pricing page, with the date we last checked it. They are not converted to our billing unit, because they do not measure the same thing.

Microlink

URL to data, screenshot, PDF and content for apps and agents.

Free tier
25 requests per day
Pro plan
US$ 49 per month for 46,000 requests

Published by the vendor, checked 2026-08-14.

microlink.io

Microlink is trademarked by their respective owners and mentioned only to describe the comparison people are searching for. No affiliation is claimed.

Questions this raises

Is PageCapture affiliated with Microlink?

No. PageCapture is an independent product with its own implementation. Microlink is a trademark of its owner and is mentioned only to describe the comparison people are searching for.

Do you return the same metadata fields?

Not field for field. We return page title, description, Open Graph and the rendered content you asked for; if your code depends on a specific enrichment field, check it against the options reference before migrating.

Is it a drop-in replacement?

No. The response envelope and the option names are ours. For an agent integration the practical path is the published skill, which carries the request shape and the retryable error codes.

What is missing today?

Region first, because enrichment usually needs it: there is no managed proxy pool behind ip_country_code, so a country-specific render is not something we serve yet. There is also no GPU pool behind request_gpu_rendering, and checkout is not wired to a merchant of record. If your pipeline depends on any of those, this is not the moment to move.