You are deciding between two scheduling tools for your team, and you have twenty open tabs trying to work out which one wins. Here is the whole thing done once, with no tab-switching: a two-product comparison study run inside Perplexity, with Cobalt connected so the assistant renders the pages itself and hands back a shareable review.
The two products in this walkthrough are Calendly and Cal.com. The decision is narrow on purpose: which one a five-person team should pay for. A comparison study answers a single question well, so pick the question before you pick the products.
Connecting Cobalt to Perplexity first
Cobalt runs as an MCP connector, the standard way an assistant talks to an outside tool. In Perplexity you add it once, paste https://cobaltcapture.com/mcp, and it stays available. The full setup lives on the Perplexity connector page, and if you also work in other assistants, the same server drops into Cursor and Claude Code without changes.
Once it is connected, you do not write code or craft a prompt. You type what you want in plain words.
The exact ask
Here is the message that kicked off the study:
Run a comparison study of calendly.com and cal.com for a 5-person team deciding which to buy. I care about per-seat price, team scheduling, round-robin routing, and the free tier.
Cobalt took it from there. The public-page studies (pricing, website, comparison) run end to end with nothing for you to do, because Cobalt renders each page rather than asking your browser to. For a comparison it also runs a web search to date the claims and catch anything the marketing pages leave out. You walk away and come back to a finished review.
What came back
The review led with the verdict, which is where you want it when you are busy. The top line was a plain lean: for a five-person team that needs round-robin routing, Cal.com worked out cheaper at this size, while Calendly's team setup read as simpler on the pricing page itself. Below that sat the evidence, so the lean was something you could check, not swallow.
The matrix was the center of it. Dated, sourced, and limited to the four things the ask named:
| Factor | Calendly | Cal.com |
|---|---|---|
| Per-seat price (team tier) | Listed per seat, billed per seat | Lower per seat at this tier |
| Team scheduling | Available on paid team plan | Available, including self-hosted |
| Round-robin routing | Paid tier only | Included lower down |
| Free tier limits | One event type, single user | Unlimited event types, single user |
Every cell carried a source and a date, because a comparison study pulls from the live pages and web search rather than from whatever the model remembered. That dating matters: pricing pages change, and a figure from eight months ago is a guess dressed as a fact.
Under the matrix came the screenshots. Cobalt judged the real pricing pages and captured them, so when the review said "round-robin is paid only" you saw the pricing page that said so. The findings were ranked by what the decision costs, not by what was easiest to spot, which is the difference between a study and a feature dump.
What the study did not do
It did not sign up, click through a trial, or test the actual booking flow. A comparison study reads public pages plus web search. If you need to know how round-robin behaves once you turn it on, that is a flow walkthrough or a usability pass on a tool you own, both of which need an assistant that can drive a browser. The comparison study is the fast, read-only first cut: it tells you which two to shortlist and where the real differences sit before you spend a trial on either.
It also stayed neutral on tone. The lean said which fit the stated need better. It did not trash either product, because the study exists to help you decide, not to sell you one side.
Sharing the result
The review published to a short link, and the markdown version at /r/<slug>/markdown is there when you want to paste the raw matrix into a doc or hand it to another assistant. If your decision-maker reads on a phone, the clean read-only page opens with no app and no login. The same output serves a person skimming for the verdict and an agent reading the structure. That dual format is the point of a product comparison you can share rather than keep in a chat log.
Comparison, pricing, and website teardowns are free to try: unlimited reviews, no password, no card. Cobalt asks for an email when the review is ready so it can send the link. After seven days a review goes read-only for you, though the link you shared keeps working for everyone you sent it to. Pro keeps everything editable for $5 a month, and you can see the full pricing before deciding.
Want the shape of the ask without the full worked example? Running a comparison study of two competitors in Perplexity covers the prompt patterns. Then connect Cobalt and ask your own two-product question.