UserProof / Free tools / CSP checker
CSP checker: will your analytics and replay scripts load? For founders who build with a coding agent
A CSP checker reads your Content-Security-Policy the way a browser does and tells you what it blocks. For founders who build with a coding agent: paste the policy below. You get three tables: whether each analytics or session replay tool would load, the policy's weak spots in order, and which directive the browser uses for each kind of request. Hand the result to Claude Code, Codex or Cursor and it can make the change.
Check a policy
Example result for the policy in the box
Read as an enforced Content-Security-Policy with 5 directives: 11 of 11 tools blocked (FullStory US and EU count once), 1 finding to fix first.
Will each tool load?
| Tool | Under this policy | What is missing |
|---|---|---|
| UserProof | Blocked | script-src needs https://ingest.getuserproof.comconnect-src needs https://ingest.getuserproof.com |
| Google Analytics 4 (Google tag) | Blocked | connect-src needs https://www.googletagmanager.comconnect-src needs https://*.google.com |
| PostHog | Blocked | script-src needs https://*.posthog.comconnect-src needs https://*.posthog.comscript-src needs blob:script-src needs data: |
| Sentry (with Session Replay) | Blocked | script-src needs https://browser.sentry-cdn.comscript-src needs https://js.sentry-cdn.comconnect-src needs *.sentry.ioscript-src needs blob:default-src needs blob: |
| Hotjar | Blocked | script-src needs https://*.hotjar.comconnect-src needs https://*.hotjar.comconnect-src needs https://*.hotjar.ioconnect-src needs wss://*.hotjar.comdefault-src needs https://*.hotjar.comdefault-src needs 'unsafe-inline' (the styles it injects) |
| Microsoft Clarity | Blocked | script-src needs https://*.clarity.msscript-src needs https://c.bing.comconnect-src needs https://*.clarity.msconnect-src needs https://c.bing.com |
| FullStory (US) | Blocked | script-src needs https://edge.fullstory.comscript-src needs https://rs.fullstory.comconnect-src needs https://edge.fullstory.comconnect-src needs https://rs.fullstory.comscript-src needs blob: |
| FullStory (EU) | Blocked | script-src needs https://edge.eu1.fullstory.comscript-src needs https://rs.eu1.fullstory.comconnect-src needs https://edge.eu1.fullstory.comconnect-src needs https://rs.eu1.fullstory.comscript-src needs blob: |
| LogRocket | Blocked | script-src needs https://cdn.logrocket.ioscript-src needs https://cdn.lr-ingest.ioscript-src needs https://cdn.lr-in.comscript-src needs https://cdn.lr-in-prod.comscript-src needs https://cdn.lr-ingest.comscript-src needs https://cdn.ingest-lr.comscript-src needs https://cdn.lr-intake.comscript-src needs https://cdn.intake-lr.comscript-src needs https://cdn.logr-ingest.comscript-src needs https://cdn.lrkt-in.comscript-src needs https://cdn.lgrckt-in.comscript-src needs https://cdn.logr-in.comconnect-src needs https://*.logrocket.ioconnect-src needs https://*.lr-ingest.ioconnect-src needs https://*.logrocket.comconnect-src needs https://*.lr-in.comconnect-src needs https://*.lr-in-prod.comconnect-src needs https://*.lr-ingest.comconnect-src needs https://*.ingest-lr.comconnect-src needs https://*.lr-intake.comconnect-src needs https://*.intake-lr.comconnect-src needs https://*.logr-ingest.comconnect-src needs https://*.lrkt-in.comconnect-src needs https://*.lgrckt-in.comconnect-src needs https://*.logr-in.comscript-src needs blob:default-src needs blob: |
| Mouseflow | Blocked | script-src needs *.mouseflow.comconnect-src needs *.mouseflow.comdefault-src needs *.mouseflow.com |
| Lucky Orange | Blocked | script-src needs https://tools.luckyorange.comconnect-src needs https://*.luckyorange.comconnect-src needs https://pubsub.googleapis.comconnect-src needs wss://*.visitors.livescript-src needs blob: |
| Smartlook | Blocked | script-src needs https://*.smartlook.comscript-src needs https://*.smartlook.cloudconnect-src needs https://*.smartlook.comconnect-src needs https://*.smartlook.cloud |
The CSP generator adds the missing sources to this policy without breaking the default-src fallback.
Weak spots
| Level | Finding | Fix |
|---|---|---|
| Fix first | 'unsafe-inline' in script-src lets any injected <script> block run, which is most of what a CSP is there to stop. | Put a nonce or hash on your inline scripts; browsers then ignore 'unsafe-inline' and it can stay for very old browsers. |
| Worth fixing | object-src falls to default-src ('self'), so plugins load from those sources. | Add object-src 'none', as web.dev's strict policy does. |
| Worth fixing | No base-uri. It has no fallback to default-src, so an injected <base> tag can point your relative script URLs at another host. | Add base-uri 'none' (or 'self' if your pages use a <base> tag). |
How the browser reads each request type
| Request | Directive used | Sources |
|---|---|---|
| Scripts | script-src | 'self' 'unsafe-inline' https://www.googletagmanager.com |
| Styles | default-src | 'self' |
| Images | img-src | 'self' data: https: |
| Fonts | default-src | 'self' |
| Fetch, beacons and sockets | connect-src | 'self' https://*.google-analytics.com |
| Frames | default-src | 'self' |
| Workers | script-src | 'self' 'unsafe-inline' https://www.googletagmanager.com |
| Plugins (object, embed) | default-src | 'self' |
How to read the result
Will each tool load? The 11 tools are UserProof, Google Analytics 4, PostHog, Sentry, Hotjar, Microsoft Clarity, FullStory (US and EU), LogRocket, Mouseflow, Lucky Orange, Smartlook. For each one, the checker takes the hosts the vendor's own CSP page lists and asks the directive the browser would use whether it allows them. An external script is judged by script-src-elem, then script-src, then default-src. A worker is judged by worker-src, then child-src, then script-src, then default-src. A frame is judged by frame-src, then child-src, then default-src, and the rest by their own directive, then default-src. A directive with nothing to fall back on allows everything. "Loads if" means the tool's inline snippet needs your nonce, or your policy uses 'strict-dynamic' and the tool must be loaded by a script that carries the nonce.
Weak spots come in three levels. "Fix first" is something that lets an injected script run or breaks the policy outright (a missing semicolon, a keyword without quotes). "Worth fixing" closes a known gap, such as base-uri, which has no fallback to default-src. "Note" is minor or informational.
How the browser reads each request type shows the directive that applies to scripts, styles, images, fonts, fetch and sockets, frames, workers and plugins, after fallback. "none set" means the browser puts no limit on that type.
The example, line by line
The box opens on a common shape: default-src 'self'; script-src 'self' 'unsafe-inline' https://www.googletagmanager.com; connect-src 'self' https://*.google-analytics.com; img-src 'self' data: https:; frame-ancestors 'none'. Google Analytics comes closest. Its script host is in script-src, the inline snippet passes on 'unsafe-inline' and img-src allows any https: host, but connect-src lacks https://www.googletagmanager.com, https://*.google.com, so the hits it sends are refused. All 11 tools are blocked, UserProof among them, because script-src and connect-src list neither its host nor a scheme that covers it. The weak spots: 'unsafe-inline' with no nonce, no base-uri, and object-src falling back to 'self'.
What 99 YC startup policies would block
59 of the 99 enforced YC startup policies would block UserProof, and 35 let all 11 tools load. On November 17, 2026 we read the home page of every company in our YC homepage benchmark (376 fetched, 371 answered) and kept the 99 that enforce a Content-Security-Policy, by header or meta tag. Then we ran this page's own check over each one. Two sites send both a header and a meta tag. A request has to pass both, so for those we count a tool as blocked when either policy blocks it, and a weak spot only when neither policy closes it.
| Tool | Blocked | Loads if (nonce) | Loads |
|---|---|---|---|
| Mouseflow | 63 | 1 | 35 |
| Sentry (with Session Replay) | 62 | 1 | 36 |
| Hotjar | 62 | 1 | 36 |
| LogRocket | 62 | 1 | 36 |
| FullStory (US) | 61 | 1 | 37 |
| Lucky Orange | 61 | 1 | 37 |
| PostHog | 60 | 3 | 36 |
| Microsoft Clarity | 60 | 1 | 38 |
| UserProof | 59 | 1 | 39 |
| Google Analytics 4 (Google tag) | 59 | 1 | 39 |
| Smartlook | 59 | 1 | 39 |
Source: UserProof, 99 enforced CSPs from YC startup homepages read November 17, 2026 (data/yc-csp-2026-11-17.json), checked by scripts/build-csp-checker.py with the rules on this page (data/yc-csp-check-2026-11-17.json). FullStory's EU row is left out; it matches the US row's shape.
35 of the 99 policies let all 11 tools load as they are, and 59 stop every one until its hosts are added (or its snippet carries a nonce). All 35 that let everything through are among the 37 with no script rule at all, which set only things like frame-ancestors or upgrade-insecure-requests. Once a site writes a script rule, every tool it adds later needs its hosts listed, and that is the moment to run the checker.
| Weak spot | Policies | Share of 99 |
|---|---|---|
| object-src is not 'none' | 44 | 44% |
| 'unsafe-inline' in the script rule, with no nonce or hash | 43 | 43% |
| No base-uri | 39 | 39% |
| No script-src or default-src at all | 37 | 37% |
| 'unsafe-eval' in the script rule | 16 | 16% |
| No frame-ancestors (header policies) | 8 | 8% |
| Scripts allowed from *, https:, http: or data: | 4 | 4% |
| A plain http source | 3 | 3% |
| A deprecated directive | 3 | 3% |
| A keyword without quotes | 1 | 1% |
| A directive written twice | 0 | 0% |
| 'none' beside other sources | 0 | 0% |
| A missing semicolon | 0 | 0% |
| 'strict-dynamic' with no nonce or hash | 0 | 0% |
| A directive a meta tag can't carry | 0 | 0% |
17 of the 99 have no "Fix first" or "Worth fixing" finding at all. The most common gap is object-src left at something other than 'none': 44 of 99. web.dev's strict policy closes the top three with three directives: script-src 'nonce-...' 'strict-dynamic'; object-src 'none'; base-uri 'none'.
What each check looks for
| Check | Level | Why it matters |
|---|---|---|
| No script-src or default-src | Fix first | Without either, the browser limits no scripts. |
| 'unsafe-inline' with no nonce or hash | Fix first | Any injected inline script runs. With a nonce or hash present, browsers ignore 'unsafe-inline' (MDN). |
| *, https:, http: or data: for scripts | Fix first | Scripts load from any host. Ignored when 'strict-dynamic' is present, since browsers then drop host and scheme sources. |
| 'strict-dynamic' with no nonce or hash | Fix first | Host sources are ignored, so nothing can start the script chain. |
| Keyword without quotes, missing semicolon | Fix first | self reads as a host called "self"; a directive name inside another list is just a host name to the browser. |
| 'unsafe-eval' | Worth fixing | Strings can run as code. |
| object-src not 'none' | Worth fixing | web.dev's strict policy sets it to 'none' to stop plugin content. |
| No base-uri | Worth fixing | No fallback to default-src; an injected <base> can redirect relative script URLs. |
| Duplicate directive, plain http source | Worth fixing | Browsers use the first copy only; http can be read or changed in transit. |
| No frame-ancestors; meta-only directives | Note | Framing by other sites; a meta tag can't carry frame-ancestors (MDN), report-uri or sandbox (CSP Level 3). |
| Report-only, deprecated directives, 'none' beside sources | Note | Nothing is blocked; the directive does nothing in current browsers; 'none' is ignored there. |
What it can't do
- It doesn't fetch your site. Paste the policy you send; the browser's network tab shows it on the page's first response.
- It checks each tool's base install, with hosts as the vendor writes them. We count a socket host the vendor writes with
wss://as allowed only when the policy listswss:,ws:, thatwss://host or*. Toolbars, surveys, chat and ad features need more hosts; the generator's reference table lists each vendor's page. - A host source with a port other than 443 counts as not covering the vendor's host.
- Under
'strict-dynamic'it doesn't decide workers: a tool whose worker falls back toscript-srcshows "Loads if". Setworker-srcto the vendor's sources to be sure. A source with a path shows "Loads if" too, because only the full script URL settles it. - It reads one policy. If you send two (two headers, or a header and a meta tag), every request must pass both.
- It doesn't grade every attack. Google's CSP Evaluator goes deeper on script-injection bypasses (for example, hosts that serve JSONP); use both.
Where UserProof fits
UserProof records real visits with one external script tag and hands your coding agent the one fix to make, with the file and line. Under a CSP it needs two sources and nothing inline: https://ingest.getuserproof.com in script-src and connect-src. Field values are never recorded. Free covers 3 sites and 5,000 recorded visits a month; Pro is $29 a month for 50,000. To add the missing hosts, use the CSP generator; to read the violation reports your policy sends, the CSP report reader; for the recorder's speed cost, see whether session replay slows a site; to keep form fields out of recordings, the masking snippet generator. All free tools are on the tools page.
Sources, read November 17, 2026
- MDN: Content-Security-Policy (fallback lists; 'unsafe-inline' ignored beside a nonce or hash; host sources ignored beside 'strict-dynamic'), worker-src (falls back to child-src, script-src, then default-src) and frame-ancestors (not supported in a meta tag).
- W3C: Content Security Policy Level 3 (report-uri, frame-ancestors and sandbox are not supported in a meta element).
- web.dev: Mitigate cross-site scripting (XSS) with a strict Content Security Policy (last updated 2024-09-13): the nonce and hash policies with object-src 'none' and base-uri 'none'.
- Google: CSP Evaluator, the deeper script-injection grader named under "What it can't do".
- Vendor hosts: each vendor's own CSP page, listed with its date on the CSP generator's sources, read November 17, 2026.
- UserProof: 99 enforced policies from YC startup homepages read November 17, 2026, data/yc-csp-2026-11-17.json.
Cite this
UserProof, "CSP checker for analytics and session replay scripts," https://getuserproof.com/csp-checker; vendor hosts from each vendor's CSP page and 99 YC startup policies checked on November 17, 2026.
FAQ
What does a CSP checker do?
It reads a Content-Security-Policy the way a browser does and tells you what the policy allows and blocks. This one also says whether 11 analytics and session replay tools would load under it, using the hosts each vendor lists on its own CSP page.
Why is my analytics script blocked by the Content Security Policy?
The script's host is missing from script-src (or default-src, when there is no script-src), or the host it sends data to is missing from connect-src. Paste your policy above: the table names the missing directive and host for each tool.
Is 'unsafe-inline' in script-src bad?
It lets any injected inline script run, which is the main attack a CSP exists to stop. A nonce or hash on your own inline scripts is the fix; browsers then ignore 'unsafe-inline'. Of the 99 YC startup homepages we checked that enforce a CSP, 62 set a script rule, and 43 of those allow 'unsafe-inline' with no nonce or hash.
Does a CSP meta tag work the same as the header?
For most directives, yes. The CSP spec says frame-ancestors, report-uri and sandbox are not supported in a meta tag, so send those as a header.
What is the difference between this and Google's CSP Evaluator?
Google's CSP Evaluator grades a policy's protection against script injection. This checker runs similar checks for the main weak spots, and adds which analytics and replay tools your policy would block, with the exact host to add for each.
Does the checker send my policy anywhere?
No. The checks run in your browser, and the form is excluded from UserProof's own recorder on this page.
What does UserProof need in a CSP?
Two sources: https://ingest.getuserproof.com in script-src and in connect-src. No inline snippet, no worker. 59 of the 99 YC policies would block it until those two are added.