A RevOps question about the market usually sounds simple. How many vendors in our category publish a free tier? Who still sells per seat? Which of them gate SSO behind the enterprise plan?
The answers sit on forty pricing pages. Someone opens them one at a time and pastes into a spreadsheet. A quarter later the sheet is stale, and nobody can say which cell came from where.
A market research API for RevOps should turn that into a script: one question, asked of every vendor's page, returning a row for each vendor with the sentence that answers it.
What kind of market research API does RevOps need?
There are two kinds, and most of the pages ranking for this phrase sell the first. Data vendors license datasets: firmographics, intent signals, market-size estimates. You query their database, and they decide what goes in it.
CrawlForge is the second kind. It owns no market data, no company records and no TAM figures. It reads public pages you point it at. That makes it the right tool when the answer is published on the vendors' own sites and no dataset has a column for it: packaging, plan limits, and which features sit behind which tier.
How do you ask one question across a whole category?
Use scrape with a question format. The hosted API runs no model for it. It ranks the page's sentences against your question and returns the five best verbatim, each with an offset into the page's markdown.
curl -X POST https://crawlforge.dev/api/v1/tools/scrape \
-H "X-API-Key: cf_live_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"url": "https://vendor-a.com/pricing",
"formats": [
"markdown",
{ "type": "question", "question": "Is SSO included on every plan or only on enterprise?" }
]
}'The answer is in formats.answer.evidence: up to five sentences, each with a score, an offset and a length. Asking for markdown in the same call costs nothing extra and lets you read the lines around each offset.
Run that over your vendor list and write one row per vendor: URL, question, top evidence, date checked. In n8n it is one HTTP Request node inside a loop; the n8n setup is in the docs.
Why not let a model write the answer?
Because nobody can audit that cell afterwards. A model can put a price on a row that the page never printed, and the fake looks exactly like a real one. The due diligence post covers that failure in detail. It matters just as much when the row feeds a deal desk's discount policy.
The extractive answer is less convenient and easier to audit. You get ranked passages, not a verdict. When we asked our own pricing page whether there is a free tier, the top sentence was "That's the free tier." It was correct, but it means nothing without the context its offset points to. Treat the evidence as a locator. A person, or a model that reads only those five sentences, fills in the cell, and the source sentence goes in the row beside it.
Two limits to plan around. The hosted API flattens tables to text, so a price that only appears in a pricing grid comes back as a run of cells rather than a clean row; the MCP server keeps table rows intact. And ask one question per call, so each row maps to one question.
What does a category survey cost?
A question format adds 1 credit to the scrape's 2, so each page costs 3.
| Survey | Pages | Questions | Credits |
|---|---|---|---|
| One question, one category | 40 | 1 | 120 |
| Pricing and packaging pass | 40 | 3 | 360 |
| That pass, quarterly for a year | 40 | 3 × 4 | 1,440 |
The free tier's one-time 1,000 credits cover two full pricing-and-packaging passes, with credits left over.
What won't it do?
It won't size a market, and it won't tell you who is buying. If the question is how many companies in EMEA have more than 200 employees, buy the dataset. It also won't watch the pages between surveys. For "tell me when a competitor changes its pricing", use hosted monitors, and per-account profiles belong in an enrichment pipeline.
Start free with 1,000 credits and run one question across your category before you build anything around it.
