What counts toward UserProof’s monthly recorded visit limit?
A recorded visit is one session from a visitor on your live site, after 30 idle minutes a new one starts. Bots dropped at ingest, your localhost dev server and setup checks do not count. All your sites share one pool, reset at 00:00 UTC on the 1st. Free holds 5,000; over the cap 1 in 4 visitors keeps recording.
This is for founders who build with a coding agent such as Claude Code, Codex or Cursor and want to know what burns the monthly allowance before they pay for it. Free is $0 for 5,000 recorded visits, Pro is $29 a month for 50,000. The rules below come from the product code, not from a general definition of a session.
One number only this page shows: the plan code (src/lib/plans.mjs, read November 9, 2026, 3 plans) caps monthly recorded visits at 5,000 on Free, 50,000 on Pro and 250,000 on Scale. Past the cap, 1 visitor in 4 keeps recording. Sources are at the bottom.
What counts and what does not?
A visit counts when ingest stores it as a prod session, and nothing else does. The table lists each case and the code that decides it.
| Event | Counts? | Why | The code that decides it |
|---|---|---|---|
| A visitor opens your live site | Yes, 1 | A new session is stored with environment prod. | resolveSession in server/ingest.mjs; the env = 'prod' filter in usage(), server/api.mjs |
| The same visitor comes back within 30 minutes of their last activity | No, same visit | The new session id continues their existing session. | SESSION_IDLE_MS (30 minutes) in resolveSession |
| The same visitor returns after more than 30 minutes idle | Yes, 1 more | No live session to continue, so a new one is stored. | resolveSession |
| A visitor on a subdomain of your site, such as go.yoursite.com | Yes | Allowed without setup, recorded with pages keyed by host. | originAllowed, subdomainOf |
| A bot on your live site: bot user agent, webdriver flag, headless-brand browser or no user agent | No | Dropped before anything is stored; only a daily drop counter goes up. | botReason in src/lib/engine/bots.mjs; countDrop in handleEvents |
| A crawler that passes every one of those checks | Yes | Stored as a normal prod session. The engine can later leave it out of findings, but the usage count does not look at that. | botReason (no match); farmSessions runs in the engine, not in usage() |
| Your local dev server: localhost, 127.x.x.x, ::1, .localhost, .local, .test | No | Stored as a dev session so setup can see the first event, and left out of the count. | LOCAL_HOST and envOf in server/ingest.mjs |
A setup check from a script tag with data-wul-env="verify" | No | Stored as a verify session. | envOf |
| A page on a host you have not allowed, such as an unlisted vercel.app preview | No | Refused with a 403 and tallied as an origin drop. | originAllowed, countDropOrigin |
| A script flooding events from one address | No past 400 batches per 10 minutes | Refused with a 429. | RATE and overRate |
| A new visitor after you pass the cap | About 1 in 4 do | The rest are dropped as over_plan; the ones kept still add to the count. | sampledIn and SAMPLE_OVER = 4 |
| Analyses, fix checks and site scans | No | They are not visits. Analyses are logged separately and unlimited; scans are free up to 10 a day per site. | analyses count in usage(); analysis_limit: null |
Read from server/ingest.mjs, server/api.mjs, src/lib/engine/bots.mjs and src/lib/plans.mjs in the main checkout, November 9, 2026.
What is one recorded visit?
One recorded visit is one stored session: a row in the sessions table with environment prod. A session starts with the first event batch from a visitor who has no live session on that site. Visitors are keyed by a visitor id from the script tag, and a new session id from the same visitor within 30 minutes of their last activity attaches to the old session instead of making a new one. The count does not wait for a click or a scroll. A person who loads one page and leaves is one visit.
The script tag decides nothing about the count. It sends events, and the ingest server applies the rules in the table. What the tag captures is on what the script tag records. Field values are never recorded.
Do bots count?
Not the ones ingest recognises. On your live domain, a request is dropped in four cases. The user agent matches the bot pattern, which covers crawlers, headless browsers, Playwright, Puppeteer, Selenium, curl, link-preview fetchers and common SEO crawlers. The browser reports a webdriver. It is a headless-brand browser. Or it sends no user agent. Nothing from it is stored.
The same checks run on a dev server or a setup check, but there the session is kept and tagged, because setup is often your agent’s own Playwright browser. It is still outside the count, because it is not prod.
The gap is a crawler that looks like a person. The bot file itself describes Linux Chrome fleets that pass every user agent check and move the mouse. Those are stored as prod and count against your allowance. The engine can leave them out of findings, but the usage count does not look at that.
Are visits pooled across my sites?
Yes, one pool covers the whole account. The usage query joins sessions to your sites and counts every one of them for the calendar month, and the cap check at ingest uses the same account-wide number. Free allows 3 sites that share 5,000 visits. Pro and Scale have unlimited sites that share 50,000 and 250,000. A quiet site does not keep a share for itself.
When does the month reset?
The count resets at 00:00 UTC on the 1st of each month. Both the usage query and the ingest check count sessions started since the first of the current month in UTC, and the reported reset time is the first instant of the next month in UTC. In New York in winter that is 7 pm on the last day of the month. Nothing carries over, and sampling stops when the new month starts.
What happens when I pass the cap?
Recording keeps going at 1 new visitor in 4. Ingest hashes the visitor id and keeps a visitor when the first byte of the hash divides evenly by 4, so the same visitor is always in or always out. A kept visit still adds to the count, so the number can run past the cap. Visits already under way when you pass it carry on. The count at ingest is cached for up to a minute, so the switch is not exact to the visit. The upgrade path itself is in how to upgrade UserProof to Pro, and plans are side by side on pricing.
How do I see my usage?
Ask your agent for your usage; it runs the get_usage tool in the UserProof MCP server, which is free. The first line reads like free plan: 4,200 of 5,000 recorded visits this month (resets 2026-12-01)., followed by sites against the site limit and the retention period. Past the cap it adds Over the limit: recording 1 in 4 visits until 2026-12-01. From 80% of the cap the agent also prints a heads-up. The number is a fresh count of prod sessions, so dev and setup sessions are not in it.
A worked example: 3 sites, a bot burst, a dev session and a setup check
This example uses made-up traffic on the real rules, on the Free plan with 5,000 recorded visits. Say your three sites get 3,400, 1,500 and 700 new real visits in November, 5,600 in all. During the month a scraper sends 900 requests as python-requests. Your agent runs 120 sessions on localhost:3000 while building, and one setup check with data-wul-env="verify".
| Traffic | Sessions | Counted |
|---|---|---|
| Real visits, three sites | 5,600 | First 5,000 in full, then about 150 of the last 600 (1 in 4) |
Bot burst (python-requests) | 900 | 0, dropped as bot_user_agent |
| Local dev on localhost:3000 | 120 | 0, stored as dev |
| Setup check | 1 | 0, stored as verify |
Total on get_usage | about 5,150 of 5,000, with sampling on until December 1 |
The pool is shared, so it makes no difference which site the visits came from. Had 300 of the 5,600 been crawler-farm sessions that slip through the bot checks, they would have used 300 of the first 5,000. To price the same traffic on other tools, use the session replay cost calculator; to see how other free plans cap or sample the same visits, use the session replay sampling calculator.
Excluded areas still count as a visit: how to keep part of your site out of recordings.
Sources, read November 9, 2026
- Plan caps,
SAMPLE_OVER = 4, calendar-month metering:src/lib/plans.mjs, UserProof source at GetUserProof/userproof, read November 9, 2026 - Session keying, 30-minute idle,
LOCAL_HOST,devandverifyenvironments, bot drops, over-cap sampling:server/ingest.mjs, read November 9, 2026 - Bot patterns and crawler-farm note:
src/lib/engine/bots.mjs, read November 9, 2026 - Usage query, pooling across sites, reset time, sampling message:
server/api.mjs(usage(),resetsAt), read November 9, 2026 - The
get_usagetool and its output:mcp-server/index.js, read November 9, 2026 - Wording quoted from the pricing FAQ and plan prices: getuserproof.com/pricing, read November 9, 2026
Comparing what a recorded visit means elsewhere? Session replay tools for coding agents, ranked covers how each one counts and charges.
FAQ
Does a bot or crawler count toward my UserProof visit limit?
Not when ingest recognises it. A user agent that matches a bot pattern, a webdriver flag, a headless-brand browser or a missing user agent is dropped on your live domain and never stored. A crawler that passes those checks is stored as a visit and does count.
Does my localhost dev server use up my recorded visits?
No. Traffic from localhost, 127.x addresses, ::1 and names ending in .localhost, .local or .test is stored as a dev session, and the usage count only includes sessions marked prod. Your live domain counts normally.
Does a visit on one site use up the limit for my other sites?
Yes. The count is per account, not per site: the usage query adds up the visits of every site you own for the calendar month. Free allows 3 sites, and all three share the same 5,000 recorded visits.
Is a visitor who comes back after lunch one visit or two?
Two, if they were idle for more than 30 minutes. A new session from the same visitor within 30 minutes of their last activity continues the existing visit. After a longer gap it starts a new session, which counts again.