Skip to content
filmrm

Layerbase app · formerly Session Replay

The film room for your product

Filmroom watches real user sessions with DOM-accurate playback, then a model reviews the tape for you: weekly reports on where people get stuck, why they do not convert, and which bugs to fix. It is an embedded app backed by a managed Postgres you own.

Flat $9/mo standalone, or included free with Pro.

filmrm/ Sessions
130 sessions
Pricing page scroll
2m 14s
Docs to signupMobile
4m 02s
Checkout drop-offErrors
38s
Read mTLS guideBot
21s

This is a real session on layerbase.com

Not a mockup. These are two actual visitor sessions captured by Filmroom, exported and embedded here. Press play and watch the DOM replay exactly as it happened, on desktop and on a phone.

Loading replay...
0:00 / 0:00

A real visitor reading /docs/cloud/connecting. Recorded with Filmroom on layerbase.com, masked and anonymized.

Paste one snippet. Watch the tape.

Filmroom is an embedded app: Layerbase runs the container, sets the variables, and points the domain for you. There is no second service to operate.

01

Add the snippet

Copy the tag for your framework (HTML, Next.js, and six more) and drop it into your app. The recorder finds itself by data-site-id.

02

Sessions stream in

Every visit is captured as DOM, not video: fast, searchable, tiny. Errors, rage clicks, and identified visitors show up on the timeline.

03

AI reads the tape

Your model annotates each session and writes a weekly report: who converted, who struggled, and exactly what to fix next. AI is optional; without a key you still watch and tag sessions yourself.

Spent hours trying to self-host PostHog? Do not worry. Layerbase runs the container, sets the variables, and points the domain, so you will be up in minutes with Filmroom, not days.

Every error links to the session that caused it

When a request fails or a component throws, Filmroom pins it to the exact session and moment it happened. Click Watch User Session on any error and you are watching the user hit it, not reconstructing it from a stack trace. It is the shortest path from a red line in your logs to the reason it went red.

filmrm/ Errors

Error: unable to fetch api.someappname.com/data

3 sessions affected

Watch User Session

Error: cannot read properties of undefined (reading 'id')

1 session affected

Watch User Session

A model watches every session, so you do not have to

You will never rewatch 130 sessions a week. Filmroom annotates each finished session with a cheap model, then a stronger model writes per-visitor stories, themes, and cited recommendations: converted, engaged, or struggling, with the exact page and control to check. It is the difference between a pile of recordings and a to-do list.

  • Per-visitor stories grounded only in what actually happened
  • Recommendations that cite the session, page, or error
  • Generate on demand or on a weekly schedule
filmrm/ Reports
this week
AI report

Signup friction on mobile blocked three high-intent visitors

72 visitors · 5 recommendations · 30 stories

AAva
CONVERTED

Reviewed pricing, moved through create, and hit repeated dead clicks on the plan selector before completing checkout on the second attempt.

UXrecommendation

Reproduce the dead clicks on the plan selector and repair any control that does not visibly respond on touch.

Analytics and error monitoring, in the same app

Filmroom is also where you see traffic sources, top pages, and front-end errors grouped into issues by fingerprint, with occurrence counts and a likely-benign classifier for the noise. One embedded app replaces the analytics tab and the error tracker you were about to add.

  • Traffic sources and top pages, no separate analytics tab
  • Errors grouped into issues by fingerprint, with hit counts
  • Bot detection and suspicious-IP review keep the numbers honest
filmrm/ Insights
Sources
google.com43%
Direct31%
duckduckgo.com8%
chatgpt.com3%
layerbase.com2%
Errors, grouped into issues

TypeError: Failed to fetch

21 hits · 18 sessions

Uncaught #418 (hydration)

7 hits · 4 sessions

Failed to load chunk

2 hits · 2 sessions

Single-tenant, on a database you own

Your sessions live in your own managed Postgres, never pooled across customers. Mask any element and its descendants with one attribute so secrets and user data never leave the page unmasked; Filmroom even auto-flags elements that look sensitive. Drop internal IPs and emails onto a do-not-record list and matching traffic is dropped before anything is stored.

  • sr-mask / data-mask hides an element; sr-block removes it entirely
  • Form inputs are masked by default
  • Sample a fraction of traffic with data-sample
filmrm/ Masking & privacy
div.card > span.valueflagged

Auto-detected: possible unmasked value. Add data-mask to hide it and everything inside.

Do-not-record list
internal IPs & emails

Compliant by construction

Session tooling touches real user data, so Filmroom is built to keep you on the right side of it: data residency, origin allowlisting, masking, opt-outs, and permissions, all under your control on infrastructure you own.

Data residency you own

Sessions live in your own managed Postgres, in the region you choose. Nothing is pooled across customers or handed to a third party.

Origin allowlisting & routing

Recording is gated to the origins you approve. Skip whole pages or subdomains with path and host rules so sensitive areas are never captured.

Masking by default

Form inputs mask automatically, and one attribute hides any element and its descendants. Auto-flagging surfaces anything that looks sensitive for review.

Do-not-record opt-out

Add internal IPs or emails to a do-not-record list and matching traffic is dropped before a single event is stored.

Role-based permissions

Invite your team as admins or view-only. View-only users watch replays but change nothing, and every dashboard sits behind your own auth.

Encrypted, HTTPS-only, single-tenant

The recorder runs only in secure contexts, data moves over TLS, and the whole instance is yours alone: yours to configure, export, or delete.

Bring your own model, or bring none

With no key, Filmroom is still a full session tool: watch every replay, filter by errors and rage clicks, and tag and classify sessions yourself. Add your own OpenAI or Anthropic key and it annotates each session and writes weekly reports automatically, billed to your provider and used for nothing else.

No API key? Filmroom still works.

Full playback, filtering, and manual classification with zero AI spend. Turn on annotation and reports whenever you add a key; the two models below are the ones we run on layerbase.com and what we recommend.

Session model

The cheap model that annotates every finished session: label, disposition, and friction scores.

OpenAIgpt-5.6-luna
Anthropicclaude-haiku-4-5

Report model

The stronger model for visitor stories and the weekly report, where reasoning across a whole journey pays off.

OpenAIgpt-5.6-terra
Anthropicclaude-sonnet-5

A monthly token budget caps spend: annotation pauses at the cap and resumes next month. Providing a key is the opt-in; leave it off and nothing is sent anywhere.

One tag, any stack

Paste it before the closing body tag on the pages you want to watch. Layerbase manages the container, write key, and domain; there is nothing else to wire up.

HTMLNext.jsNuxtSvelteKitAstroRemixSPA (Vite/CRA)Electron
index.html
<script src="https://replay.layerbase.dev/v1/<recorder>.js"
        data-site-id="web"
        data-write-key="wk_live_..."
        data-ingest="https://replay.layerbase.dev"
        async></script>

Identify visitors with data-user-id / -name / -email, or skip pages and subdomains with data-deny. Records over HTTPS only.

A polished mobile PWA

Install Filmroom to your home screen and review the tape from your phone: a native-style bottom nav, card layouts, and bottom sheets built for touch, not a desktop table crammed onto a small screen.

Get in before filmroom.dev

Filmroom will grow up into a standalone product at filmroom.dev, multi-tenant and with the models included, no key of your own required. Running it on Layerbase today is early access: the full app, on your own database, with bring-your-own-key while we get there.

Two ways to run Filmroom

Subscribe to Filmroom on its own, or get it free as part of a Pro plan that also runs your databases. Either way there is no per-session meter; AI insights are optional and billed to your own model provider.

Filmroom standalone

$9/mo

The full app on its own managed Postgres. Watch and classify sessions with no AI at all, or bring your own key.

  • DOM playback, insights, and grouped errors
  • Privacy masking and do-not-record, single-tenant
  • Optional AI: bring your own OpenAI or Anthropic key
  • Flat price, no per-session meter, no second vendor
Subscribe to Filmroom

Layerbase Pro

Filmroom included free

$15/mo

A full database platform with Filmroom bundled in at no extra charge. The better deal the moment you also need a database.

  • Up to 10 databases, every engine supported
  • Always-on, never archived, full-featured
  • mTLS, instant branching, 30-day backups
  • Filmroom included, plus Secret Store
Subscribe to Pro

Filmroom questions

What happened to Session Replay?

Same app, new name. Session Replay is now Filmroom: the film room where you review the tape on real visitors. The install snippet, your data, and this page all keep working; only the branding changed.

Is my data private?

Yes. Filmroom runs single-tenant on your account, backed by your own managed Postgres, with masking on any element you choose and a do-not-record opt-out list. Your sessions are yours, not pooled across other customers.

What models does it use, and who pays for them?

Bring your own Anthropic or OpenAI key, billed to you and used only for Filmroom. A cheap session model annotates every session (we recommend gpt-5.6-luna, or claude-haiku-4-5 on Anthropic) and a stronger report model writes visitor stories and weekly reports (gpt-5.6-terra, or claude-sonnet-5). A monthly token budget caps spend.

How is it different from PostHog or FullStory?

Self-hosted-style ownership without the self-hosting work: the data lives in a database you own, on a flat $9/mo with no per-session billing, and the AI reports run on your own model key. Using it on Layerbase now is early access to what becomes filmroom.dev.

filmrm

Stop guessing why they left

Add the snippet, watch real sessions, and classify them yourself or let a model do it. Two ways to start.