Session replay: what it records, what it costs, and how to use it, for founders of a small site or SaaS who build with a coding agent
Session replay is a script on your page that records its structure and every change to it, plus clicks and scrolls, then rebuilds the visit on playback. It is not a video, so it stays small and can leave out private fields. For founders of a small site or SaaS, it shows why a visitor left, which analytics cannot show.
This page maps the topic and sends you to the page that goes deep on each part. The price numbers come from UserProof's pricing ledger, read daily from 11 vendors' pricing pages, as of 2026-11-07. It shows that 7 of 11 tools print a price that records every session at 10,000 sessions a month, 5 of 11 at 50,000, and 5 of 11 at 100,000.
Which session replay tool should you pick?
Find your situation, then follow the link to that page. Prices were read November 7, 2026.
| What are you doing? | Pick | Why |
|---|---|---|
| Watching sessions at high volume and paying nothing | Microsoft Clarity | Free, no sampling, and its pricing page states no traffic limit; its FAQ caps recording at 100,000 sessions per project per day. It keeps recordings 30 days. |
| Having your coding agent get one fix with the file and line | UserProof ours | Free for 5,000 recorded visits a month, $29 for 50,000. No heatmaps or surveys. |
| Wanting replay next to analytics and feature flags | PostHog | 5,000 recordings a month free, then priced per recording. |
| Leaving Hotjar | Hotjar alternatives | Hotjar's Free plan records a 5% sample of sessions; its prices are on our Hotjar pricing page. |
| Choosing between two free-to-start tools | Clarity vs Hotjar | Clarity records up to 100,000 sessions per project a day; Hotjar's Free plan records a 5% sample. |
| Wanting the largest capped free plan | FullStory | 30,000 sessions a month free, paid plans quote-only. |
| Debugging front-end errors | LogRocket | Our LogRocket page picks it for network and console detail when you debug front-end errors. From $83 a month at 10,000 sessions, billed annually. |
| Wanting to host it yourself | OpenReplay | Free only if you host it; no hosted plan prints a price for a session volume. |
| Already paying for Sentry or Amplitude | Sentry or Amplitude | Both count replays, not sessions, so compare with care. |
Jump to: how it works · what it records · what it costs · dead and rage clicks · with a coding agent · which tool · where others win · FAQ · sources
How does session replay work?
A script on your page takes a snapshot of the page's structure when the visit starts, then logs each change to that structure and each user action, and playback applies those changes to rebuild what the visitor saw. Mouseflow's explainer, read November 7, 2026, describes the three parts: the page's structure (the DOM), the changes to it during the visit (DOM mutations), and the visitor's actions on page elements, plus some metadata. New Relic's guide, read the same day, says session replay does not literally create a video of the visit.
That design has two practical effects. A log of changes is much smaller than a video, and a script that sees each element can skip the ones you tell it to skip. Vendors differ on how they describe it: Quantum Metric's glossary calls a replay "like watching a video of a user's session", which describes what you see on playback. That opening snapshot is also where most of the speed cost sits; we weighed 10 recorder scripts and timed two on one page in does session replay slow down my website.
Replay answers a different question from analytics. Analytics tells you which step loses people, and our funnel calculator ranks the steps. The replay of people who left at that step shows what happened there. We walk through that loop, with one site's before and after, in why users leave your website. To see the setup itself, read record real visitor sessions with a script tag.
What does session replay record, and what does it mask?
It records what a visitor sees and does on the page: page loads, clicks, scrolls, and changes to the page. What it masks is set by the tool and by you. New Relic says its replay can be tuned to mask personally identifiable information and lets you choose what to hide on a page, as read November 7, 2026. Check any tool's defaults before you install it, and test with a throwaway form.
For UserProof, per its capabilities and privacy pages read November 7, 2026: field values are never recorded, and it sets no cookies. A click keeps the first 50 characters of the clicked element's text, so owners mark rich-text editors with data-exclude. Add it to any element that should stay out of a recording, or list selectors in the script tag's own data-exclude attribute. Our own tool pages follow that rule: the forms on the funnel calculator carry data-exclude. The setup details are on the script tag page.
How much does session replay cost at 10,000, 50,000 and 100,000 sessions a month?
From $0 to $839 a month among the tools that print a price. UserProof's pricing ledger, read daily from 11 vendors' pricing pages, as of 2026-11-07, covers 28 plans. Each cell below is the cheapest printed plan that records every session at that volume, in the vendor's own currency. "None" means the vendor prints no such price: the plan is quote-only, drawn only by a browser-side slider, sampled, event-priced, or free only when self-hosted.
| Tool | 10,000 | 50,000 | 100,000 |
|---|---|---|---|
| UserProof ours | $29 | $29 | $99 |
| Microsoft Clarity | $0 | $0 | $0 |
| PostHog | $25 | $172.50 | $272.50 |
| Amplitude | $0 | none | none |
| FullStory | $0 | none | none |
| Lucky Orange | $72 | $199 | $839 |
| LogRocket | $83 | $333 | $717 |
| Mouseflow | €109 | €319 | €319 |
| Hotjar (Contentsquare) | none | none | none |
| OpenReplay | none | none | none |
| Sentry | none | none | none |
| Harvv | none | none | none |
Four things to know before you compare. Units differ: Amplitude and Sentry count replays, PostHog counts recordings and UserProof counts recorded visits. LogRocket, Lucky Orange and Mouseflow cells are monthly equivalents on annual billing, and Mouseflow's are in euros, not converted. PostHog's figures are our arithmetic from its published tiers. Hotjar's printed plans record a sample (5% on Free, 15% on Growth, per the ledger read November 7, 2026), so the ledger leaves them out.
The ledger's own page lets you type any volume: the session replay cost calculator. It sits with our other no-sign-up tools on the free tools page. For one vendor in full, see PostHog pricing, Hotjar pricing and Microsoft Clarity pricing. UserProof's own plans, from its pricing page: Free covers 3 sites and 5,000 recorded visits a month, Pro is $29 a month for 50,000, and Scale is $99 a month for 250,000. Past a plan's visits, UserProof keeps recording 1 in 4 visits until the month resets.
What are dead clicks and rage clicks?
A dead click is a click that changes nothing. A rage click is a burst of fast clicks on one element, usually because nothing is happening. Both point to a specific element to fix, which is why they are the first thing worth searching for in a pile of recordings. UserProof counts a dead click when nothing happened within 2.5 seconds, and a rage click as 3 or more clicks on one element, each less than 1 second apart. Fixed rules count them from recorded events, never an LLM; the rules are on the dead and rage clicks page.
They are common. In our scan of 493 YC startup homepages on October 28, 2026, 33 (6.7%) had at least one dead click, and 43 of the 49 dead clicks were on real buttons and links. The method is in the dead click report. To test your own page without installing anything, use the dead click finder.
How do you use session replay with a coding agent?
Give the agent the finding, not the video. A founder who builds with Claude Code, Codex or Cursor rarely wants to watch hours of recordings; the agent needs the page, the element and the file to change. There are two routes. One is to let the agent query a replay tool you already run: Clarity's MCP server in Claude Code works this way, within the 10 Data Export API requests a project a day that Microsoft's docs allow. The other is a tool built to hand the agent a ranked fix.
UserProof is the second kind. It works in Claude Code, Codex and Cursor, and any agent that runs MCP servers; it gives the agent the one highest-impact fix with the file and line to change, then checks whether your goal moved. Setup is one terminal command: set it up in Claude Code or install it in Codex or Cursor. After you ship, check whether the fix worked. For a ranked comparison of 12 tools on six stated criteria, see session replay tools for AI coding agents, and for PostHog's side, PostHog MCP vs UserProof.
How do you choose between the tools?
Decide on three things in this order: the volume you will record, whether you need every session or a sample, and who reads the results, you or your agent. The pages below do the one-to-one work.
- Leaving a tool: Hotjar, Microsoft Clarity, PostHog, LogRocket, FullStory, Sentry, Amplitude and OpenReplay alternatives.
- Pricing in full: PostHog, Hotjar and Microsoft Clarity.
- Two tools side by side: Clarity vs Hotjar, UserProof vs Clarity, LogRocket vs PostHog, Amplitude vs PostHog and PostHog MCP vs UserProof.
Where do other tools still win?
Several places, and you should weigh them. Microsoft Clarity is free, states no traffic limit and includes heatmaps, which UserProof does not have, though its FAQ says it records up to 100,000 sessions per project per day; both pages were read November 7, 2026. FullStory's free plan covers 30,000 sessions a month with 12 months of retention for up to 10 users, against UserProof Free's 5,000 visits and 7 days; FullStory's paid plans are quote-only, as read on its pricing page. PostHog puts replay beside analytics, feature flags and surveys, which UserProof does not offer. If you need a heatmap, a survey or an analyst dashboard, pick one of those tools.
FAQ
What is the main purpose of session replay?
To show why a visitor left, not just that they did. Analytics counts the drop at a step; a replay rebuilds what that visitor saw and did, such as the click that did nothing, the button below the first phone screen or the error that ended the visit. You then fix that one thing and check the same step again.
How much does PostHog charge for session replay?
The first 5,000 recordings a month are free. After that PostHog charges per recording: $0.005 each from 5,001 to 15,000, $0.0035 each from 15,001 to 50,000, and $0.002 each from 50,001 to 150,000. By our arithmetic that is $25 at 10,000 recordings, $172.50 at 50,000 and $272.50 at 100,000, read November 7, 2026. The full tier list is on our PostHog pricing page.
What is a session in a website?
One visitor's visit: the pages they open and what they do on them, from arrival until they leave or go idle. Each tool decides where a session ends and what to call the unit, so the ledger above counts sessions, recordings, replays or recorded visits depending on the vendor.
Does session replay need a cookie banner?
That depends on the tool and where your visitors are, so check its privacy page. UserProof sets no cookies and never records field values, as of November 7, 2026. This is a product fact, not legal advice.
Sources, read November 7, 2026
- UserProof pricing ledger and cost calculator: 11 vendors, 28 plans, prices read November 7, 2026. Vendor pages behind it: Microsoft Clarity pricing and its FAQ, PostHog pricing, FullStory pricing, Hotjar (Contentsquare) pricing, LogRocket pricing, Amplitude pricing, OpenReplay pricing, Mouseflow pricing, Lucky Orange pricing, Sentry pricing and Harvv pricing
- Mouseflow, How Session Replays Work: the page structure, the changes to it and the visitor's actions are what gets captured
- New Relic, What is session replay: not literally a video; PII masking can be tuned
- Quantum Metric, Session Replay glossary: the "like watching a video" description
- UserProof pricing: Free, Pro and Scale plans, and the 1-in-4 recording past the limit
- Script tag page and privacy page: field values are never recorded, no cookies,
data-exclude, 50 characters of click text - Dead and rage click rules, dead click report (493 YC startup homepages, October 28, 2026) and Clarity MCP guide and Microsoft's Clarity Data Export API docs (10 requests a project a day)