You have a pricing page you are not sure about. Maybe the plan tiers blur together, maybe a buyer cannot tell which one is theirs, maybe you suspect the phone view hides the price. You do not have a researcher on staff to check. What you do have is an AI assistant you already use. Here is how to hand it a URL and get back a review you can act on, without writing a prompt from scratch or touching any code.
Step 1: Connect Cobalt to the AI you already use
Cobalt runs inside your assistant over MCP, the standard way tools plug into AI. In Claude chat, open Settings, go to Connectors, click Add custom connector, and paste https://cobaltcapture.com/mcp. That is the whole setup. If you work in Claude Code instead, run claude mcp add --scope user --transport http cobalt https://cobaltcapture.com/mcp. The same connector works in Cursor, Codex, and Perplexity.
Outcome: your assistant can now run Cobalt's ready-made studies. You only do this once. If you want the exact steps for other tools, the guide on connecting your coding agent over MCP covers each one.
Step 2: Ask for the study in plain words
Type what you want. For example: "Run a pricing teardown of yoursite.com/pricing". No code, no prompt engineering. If you are not sure what studies exist, type /workflows and the assistant lists every one and exactly what to say.
A pricing page teardown answers one question: can a buyer tell what they would pay and which plan is theirs. It works out the real cost for a solo user, a team of 5, and a company of 25. It checks the fine print, the annual-versus-monthly math, and the phone view. It only reads public pages, so there is nothing to log into and nothing to share with it.
Outcome: the assistant starts the study. Because the pricing teardown runs on public pages, there is no browser to drive and nothing for you to click. Cobalt renders your pages itself and runs start to finish. You can close the tab and come back.
Step 3: Read the verdict first, then the ranked findings
When it finishes, you get a shareable review, not a wall of chat text. The verdict sits at the top: the single clearest read on whether a cold buyer can price your product. Below it, the findings are ranked by what they cost you, worst first. Each finding comes with a real screenshot of the exact page it is judging, so you are not guessing which element it means.
Read the verdict, then scan the top three findings. That is usually where the money is. A common one: the solo price is clear but the team-of-5 total requires math the page never shows, so a buyer in that bracket bounces. Another: the phone view collapses the plan comparison into a scroll that buries the price below the fold.
Outcome: a prioritized list you can send straight to whoever owns the page. If you want a sense of what makes these land, the piece on pricing teardown mistakes a founder won't read past shows the findings worth leading with.
Step 4: Share it with the person who fixes it
The review has its own short link. Anyone you send it to can open it in a browser, on a laptop or a phone, with no app and no password. If a copywriter or a developer needs to comment, they can comment on any individual finding right on the page, and you can mark each one resolved as it gets handled.
If the fix goes to an AI coding agent instead of a person, the same review is available as clean markdown that the agent reads directly, screenshots included. One review, two audiences. The write-up on one review serving a client and a developer at once walks through that split.
Step 5: Re-run it after the fix to confirm it stuck
Once the page changes, ask for the same study again: "Run a pricing teardown of yoursite.com/pricing". A re-run leads with what got fixed, so you can confirm the top findings closed instead of re-reading the whole thing. This is the fastest way to close the loop: change, re-run, verify.
Outcome: proof that the change worked, in the same format as the original, ready to share again.
What this costs and where it stops
Running studies and making reviews is free to try, with no password and no credit card. Cobalt asks for an email only when a review is ready, so it can send you the link. After 7 days a review goes read-only for you: editing, exports, and agent access stop, though the link you already shared keeps working for everyone it was sent to. If you want reviews to stay editable with no cutoff, Pro is $5 a month or $50 a year and reopens read-only reviews instantly. You can cancel any time. The full terms live on the plans and pricing page.
The pricing teardown is one of several ready-made studies. A positioning audit of your homepage checks what a cold visitor sees first, and a side-by-side product comparison stacks you against one or two rivals for a single decision. Same setup, same plain-words request, same ranked review at the end.
Next step: paste the connector URL into your assistant and ask it to run a pricing teardown on your own page. Read the verdict, fix the top finding, re-run it.