A Fortune Global 500 electronics manufacturer
Runs web data programs across its markets on Import.io.
AI & Technology web data
Train on it. Ground on it. Compete on it. The web as infrastructure for technology companies.
Import.io collects public pages, documentation, pricing and changelogs for training, retrieval and agent workflows. AI and technology teams receive structured data with provenance through the extraction API, Web Scraper MCP or a managed feed.
214competitor pages watched
9material changes this week
41.8Btokens in the latest corpus build
10Kfree MCP calls to start
| Company | Page | Change | Detected |
|---|---|---|---|
| Acme Cloud | Pricing | Pro plan $49 → $59 per seat | 16:14 |
| Acme Cloud | Pricing | SSO moved from Pro to Business | 16:14 |
| Vectorly | Changelog | Released: batch embeddings API | 15:02 |
| Northwind AI | Docs | New endpoint: /v2/agents | 14:47 |
| Relaywise | Integrations | Listed in 3 new app marketplaces | 11:20 |
| Vectorly | Careers | +12 roles: GPU infrastructure | 09:05 |
live events
10Kfree successful MCP calls, then $0.0002 each
60 sfastest monitoring cadence on changing pages
190locales and scripts, from ko to pt-BR to ar
100%of corpus documents with provenance
Technology companies already on it.
Production examples from organisations using Import.io today and in recent years.
An AI company building models for entertainment
Has fed its models with structured web data from Import.io.
A creator-economy data platform
Collects the public web data behind its API on Import.io.
Three jobs, three failure modes.
For a technology company the web does three jobs at once. It is training data for your models, the runtime your agents act in, and the market where competitors change pricing and ship features every week. Each one fails differently. Import.io is built for all three.
Corpora fail on what you can’t see: duplicates, boilerplate, missing licence signals and pages that asked not to be used. Agents fail on page text that looks like data and isn’t, and on sites that block them. Competitive intelligence fails on noise — A/B tests and personalised pricing pages that change for no one but the crawler.
Import.io answers each with the same discipline: provenance on every document, typed output against a schema, and change detection that separates a real price change from a test.
who uses it
- AI labs and model builders
- Agent and application teams
- Product marketing and competitive intelligence
- Developer relations
- Corporate development
What AI & technology teams do with it.
Six programs we run most often in this industry.
Related guidance: AI and LLM data workflows, MCP, Firecrawl and Playwright comparison, Import.io and Firecrawl comparison, training data.
Training corpora and evaluation sets
Domain corpora built to spec — deduplicated, quality-filtered, licence- and opt-out-aware — plus held-out evaluation sets.
corporaAgent web access through MCP
Search, fetch and extract as tools for Claude, Cursor and any MCP client, returning typed records rather than page text.
agentsPricing and packaging intelligence
Competitor pricing pages diffed daily or faster: plan prices, seat minimums, feature moves between tiers.
pricingChangelogs, docs and releases
New endpoints, deprecations and launches detected as they publish, not when the blog post lands.
productMarketplaces and integrations
App-store and integration-directory listings, ratings and reviews across the ecosystems you sell into.
ecosystemsHiring and developer signals
Roles, locations and skills from careers pages, plus community questions and sentiment on public forums.
signalsThe data, field by field.
| field | type | example |
|---|---|---|
| doc_id | str | sha256:1a9e… |
| url | url | docs.example.com/v2/agents |
| domain | str | docs.example.com |
| lang | str | en |
| text | str | main content, boilerplate removed |
| tokens | int | 2,418 |
| licence_signal | str | CC-BY-4.0 |
| robots | enum | allowed |
| change_type | enum | added_endpoint |
| fetched_at | ts | 2026-09-25T14:47:10Z |
where it comes from
- Documentation sites and API references
- Pricing pages and changelogs
- App stores and integration marketplaces
- Public forums and Q&A sites
- Careers pages
- News and technical publications
maintained datasets
The hard parts, handled.
What breaks when this is done with scripts, and how Import.io handles it.
01
Corpora you can defend
Near-duplicate removal, quality filtering and language identification, with licence signals, robots status and a content hash recorded for every document.
02
Output agents can use
Extraction to a schema you define, so agents act on typed fields, not on paragraphs that happen to contain a price.
03
Signal, not noise
Diffs are checked across repeated captures and regions, so A/B tests and personalisation don’t show up as competitor moves.
04
Reach without drama
Rendering, challenges and regional egress are handled server-side, and blocked calls are not billed.
How a program is scoped.
A clear operating plan before collection begins.
01
Sources
Agree the public sites, pages and coverage the program needs.
02
Schema and fields
Define the entities, fields and validation rules for each structured record.
03
Delivery cadence
Set the refresh schedule and choose API or managed-feed delivery.
Three ways to get it.
Same capture engine underneath each one.
Web data for AI
Model builders who need corpora, retrieval feeds or evaluation sets built to spec.
learn more →Web Scraper MCP
Agent teams who want structured web access in Claude, Cursor or their own MCP client, 10,000 calls free.
learn more →Self-service platform
Product marketing and CI teams who want to set up their own page monitors and extractors.
learn more →data scope
Collection honours robots.txt and AI-crawler rules, records licence signals per document, and detects and removes personal data. What a corpus may be used for is agreed per program.
AI & Technology questions.
Straight answers.
Can you build a training corpus to our specification?
Yes: domains, languages, date ranges, exclusions, deduplication and quality thresholds, delivered with provenance and a quality report per build.
How do agents use Import.io?
Through the Web Scraper MCP server or the API. Agents call extraction tools and get typed records back; the first 10,000 successful calls are free.
How do you avoid false alarms from A/B tests on pricing pages?
Changes are confirmed across repeated captures and regions before they are reported, so a test variant shown to one visitor doesn’t become a price change.
Do you respect opt-outs and licences?
Yes. Robots.txt and AI-crawler rules are honoured, and licence signals are recorded per document so your use can follow them.
Can we monitor competitors ourselves?
Yes. The self-service platform lets your team set up page monitors and extractors, from the free trial upwards.