Open-Source vs SaaS Session Replay Tools in 2027
Session replay is turning into an observability signal, a popular open-source tool lost its hosted product, and self-hosting got both easier and harder. Here is how to choose for next year.
Dmytro Vavreniuk
Product Engineer
Short version: almost every open-source session replay tool records with rrweb; they differ in what they build around it. Pick OpenReplay for self-hosted debugging, ClickStack if you want replays next to OpenTelemetry traces and logs, Sentry if you already self-host it, Umami for light replay on a marketing site. Choose SaaS when your volume is moderate and nobody wants to run ClickHouse at 3 a.m. Whatever you pick, use a standard session.id and keep an export path, because 2026 showed that open-source code does not guarantee the product survives.
The open-source vs SaaS question used to be about price. For 2027 it is about three things that changed this year. First, Highlight.io, one of the best-known open-source monitoring and replay tools, was acquired by LaunchDarkly in April 2025 and its standalone hosted service shut down on February 28, 2026. Second, session replay is moving into OpenTelemetry-based observability stacks: on October 30, 2026 ClickStack merged support for the OpenTelemetry session.id in its replay views, and in October Grafana opened its Session Replay public preview to everyone. Third, replay is spreading into lightweight tools: the privacy-first analytics project Umami added replays in version 3.1.
This guide compares the open-source options, explains what self-hosting really costs, and gives a simple way to decide. If you are new to the topic, start with our hub article what is session replay. For a pricing-focused comparison of commercial tools, see session replay tools compared.
Under the hood, almost everyone uses rrweb
Most session replay tools, open-source or not, do not record video. They capture a snapshot of the page's DOM and then every change to it, plus mouse movement, scrolling and input events, and rebuild the page in the player. The MIT-licensed rrweb library does this for PostHog, Sentry, Umami, ClickStack's browser SDK and many others. (We explain the mechanics in session replay architecture.)
That has a practical consequence: the recorder is rarely what separates tools. What differs is everything after the browser: ingest at scale, storage, search ("show me sessions with a 500 on checkout in Safari"), privacy controls, links to errors and backend logs, and who keeps it running. Here is how small the recording part can be:
// The smallest possible self-hosted recorder: rrweb + your own endpoint
import { record } from 'rrweb';
const sessionId = getOrCreateSessionId(); // e.g. a UUID kept in sessionStorage
let buffer = [];
record({
emit(event) { buffer.push(event); },
maskAllInputs: true, // never record what users type
maskTextSelector: '[data-private]', // mask marked text on the client
});
setInterval(() => {
if (!buffer.length) return;
navigator.sendBeacon('/replay', JSON.stringify({ sessionId, events: buffer }));
buffer = [];
}, 10_000);Twenty lines get events to your server. Turning them into a searchable, privacy-safe, always-available replay product is the other 95% of the work, and it is what you are choosing between.
The open-source and self-hostable options in 2027
| Tool | License | What it is | Self-hosting | Best for |
|---|---|---|---|---|
| rrweb | MIT | Recording library | You build ingest, storage and search yourself | Teams that want full control or are building a product on top of replay |
| OpenReplay | AGPLv3 core + commercial "ee" parts | Full replay + product analytics platform | Docker Compose on one server or Kubernetes; recordings in S3-compatible storage | Self-hosted LogRocket-style debugging with devtools data |
| PostHog | MIT core, separate enterprise license | Product analytics suite with replay | "Hobby" Docker deploy only, aimed at roughly 1k recordings a month; no new Kubernetes installs | Its cloud version; self-hosting is for evaluation and small sites |
| Sentry | FSL (source-available, turns Apache 2.0 after two years) | Error monitoring with replay | Included in self-hosted Sentry; heavy stack to operate | Teams already self-hosting Sentry for errors |
| ClickStack (HyperDX) | Open source | OpenTelemetry observability on ClickHouse with replay | ClickHouse + collector + UI; replays stored in a ClickHouse table | Teams that want replays next to traces and logs |
| Umami | MIT | Privacy-first web analytics, replays since v3.1 | Small Node.js + Postgres app | Marketing sites that need occasional replays, not deep debugging |
| Highlight.io | Apache 2.0 (code) | Former full-stack monitoring | Code still public, but the product is no longer actively run as Highlight | A lesson, not a choice (see below) |
Licenses and deployment options as published by each project in November 2026. Licenses change; check the repository before you commit.
OpenReplay: the most complete self-hosted replay product
OpenReplay is the closest thing to a self-hosted LogRocket: replays with console logs, network requests and performance data, plus funnels, heatmaps and, in 2026 releases, reusable segments, actions and user profiles. It installs on a single server with Docker Compose or on Kubernetes and stores recordings in S3-compatible storage. The core moved to AGPLv3 in 2024, with some enterprise features under a commercial license. If you modify it and offer it to others as a service, AGPL obligations apply; for internal use it is rarely an issue.
PostHog: great product, limited self-hosting
PostHog is open source, but its self-hosted option is a single-server "hobby" deployment recommended for small volumes (around 300,000 events and 1,000 recordings a month) and new Kubernetes deployments are not supported. In practice, PostHog is an open-source project you use as SaaS. That is fine, just do not plan a production self-hosted rollout around it.
Sentry: replay as part of error monitoring
Sentry's replay shows the user session around an error, and it is included in self-hosted Sentry. Sentry uses the Functional Source License: you can run and modify it, but not offer a competing service, and each version becomes Apache 2.0 after two years. Self-hosted Sentry is a large stack (Kafka, ClickHouse, Postgres, Redis, Snuba and more), so it makes sense mainly if you run it already. For how replay and errors work together, see error tracking with session replay.
ClickStack: replay as an OpenTelemetry signal
ClickStack is ClickHouse's open-source observability stack, built on HyperDX (which ClickHouse acquired in 2025). Its browser SDK extends the OpenTelemetry JavaScript SDK, records replays alongside traces, and propagates the traceparent header so a failed request in a replay opens the backend trace, its logs and database queries. Replays live in their own ClickHouse table. If your logs and traces are already in ClickHouse, this is the most integrated self-hosted option. The trade-off: it is a debugging tool first, with less product analytics than OpenReplay or PostHog.
Umami and other analytics-first tools
Umami 3.1 added rrweb-based replays with configurable masking to its MIT-licensed analytics; OpenPanel offers similar features. These are good when a marketing team wants to watch a few sessions on a landing page without adding another vendor. They do not capture console errors or network calls the way debugging tools do.
rrweb on its own
Building on rrweb directly makes sense when replay is part of your own product, or when you have unusual constraints (air-gapped networks, custom storage). Budget for the parts the snippet above skips: chunking and compression, retries, a player with search, retention jobs and masking reviews.
What SaaS still does better
Commercial tools (LogRocket, FullStory, Contentsquare/Hotjar, Microsoft Clarity, Datadog, Grafana Cloud and LogNroll among them) are not just hosted rrweb. You pay for things that are tedious to build: search across millions of sessions, automatic detection of rage clicks and errors, AI summaries, mobile SDKs, uptime, scaling during traffic spikes and someone else on call. For most teams below a few hundred thousand sessions a month, that is cheaper than running it yourself. Our comparison of 12 session replay tools covers prices and free tiers.
The main reasons to avoid SaaS are data residency (recordings must stay in a specific country or network), regulated industries, very high volume where per-session pricing becomes painful, and a strong in-house platform team that already runs ClickHouse or Kubernetes.
The real cost of self-hosting
Storage is rarely the problem. A compressed recording is typically a few hundred kilobytes, though long sessions on complex pages can reach several megabytes. A rough monthly estimate:
- Storage: 100,000 sessions × 300 KB ≈ 30 GB a month. In object storage that is around a dollar per month of data kept. Even with 90-day retention, it stays in single digits.
- Compute: ingest, processing and the search database (Postgres, ClickHouse, Kafka or Redis, depending on the tool). A single server with 4 to 8 vCPUs and 16 to 32 GB RAM handles a small production load; high availability doubles it. Expect roughly $100 to $400 a month in cloud costs.
- People: upgrades, backups, monitoring the monitoring, scaling during spikes, security patches and access reviews. Even 4 to 8 hours a month of an engineer's time costs more than the servers.
So the honest comparison is your SaaS quote against servers + engineer time, not against storage. Self-hosting usually wins only at high volume, when compliance rules out SaaS, or when the platform already exists for other reasons. Do not forget what to collect and what to ignore: sampling and retention limits cut both bills.
The 2027 trend: replay joins OpenTelemetry
The biggest change for next year is not a tool, it is a standard. OpenTelemetry's semantic conventions define a session.id attribute and session.start/session.end events (still in development status). Vendors are now using that ID as the join key between a replay and everything else: ClickStack reads it since October 2026, Elastic is designing replay for its OpenTelemetry browser agent, and Grafana's replay is built on its Faro SDK. Replay is no longer a separate product; it is becoming one more signal next to traces, metrics and logs.
You can prepare for this today, whatever tool you use. Generate one session ID in the browser, give it to your replay recorder, and stamp it on every span:
// Stamp the same session ID on every browser span (OpenTelemetry JS)
import { WebTracerProvider, BatchSpanProcessor } from '@opentelemetry/sdk-trace-web';
import { OTLPTraceExporter } from '@opentelemetry/exporter-trace-otlp-http';
const sessionId = getOrCreateSessionId(); // the ID your replay recorder also uses
const sessionProcessor = {
onStart(span) { span.setAttribute('session.id', sessionId); },
onEnd() {},
forceFlush() { return Promise.resolve(); },
shutdown() { return Promise.resolve(); },
};
const provider = new WebTracerProvider({
// the session processor must run before the exporter
spanProcessors: [sessionProcessor, new BatchSpanProcessor(new OTLPTraceExporter())],
});
provider.register();Send the same ID to your API (as a header or baggage) and add it to server-side logs. Then a slow trace leads to the replay, and a replay leads to the backend logs. We cover the backend half in network debugging with session replay.
The Highlight lesson: plan your exit
Highlight's code is still on GitHub under Apache 2.0, yet teams that relied on its hosted service had to migrate when it closed. Open source protects you from a vendor raising prices; it does not protect you from a project losing its maintainers. Whatever you choose, keep these escape routes:
- Own the session ID. Generate it yourself (or use the OpenTelemetry one) and store it in your logs, so links survive a tool change.
- Wrap the SDK. Call the replay SDK through one small module of your own. Swapping vendors then touches one file, not every page.
- Check exports. Can you export recordings (ideally as rrweb events) and metadata through an API? Test it before you need it.
- Watch project health. Release frequency, number of active maintainers and who funds the project matter more than GitHub stars.
- Keep retention short. If you only keep 30 days of replays, a migration means waiting 30 days, not moving terabytes.
Privacy does not depend on the license
Self-hosting keeps recordings on your servers, which simplifies data residency. It does not change your obligations: you still need a legal basis or consent, client-side masking of personal data and a retention policy. Check that masking happens in the browser and is on by default; Grafana, for example, masks all text and inputs by default. Our guides on session replay and GDPR and privacy-first session recording in the EU go into detail.
How to decide in five questions
- Must recordings stay on your infrastructure? Yes: self-host (OpenReplay, ClickStack, Sentry) or a vendor with a self-hosted or on-premises option.
- Do you already run ClickHouse, Sentry or an OpenTelemetry collector? Yes: add replay to that stack before adopting a new one.
- More than a few hundred thousand sessions a month? Get SaaS quotes and compare them with servers plus engineer time.
- Is the main user a developer or a product manager? Developers need errors, network and traces; product teams need funnels, heatmaps and search.
- Who is on call for it? If the honest answer is "nobody", choose a managed service.
Where LogNroll fits
LogNroll is a managed session replay service with console errors, network calls and click heatmaps in every plan, and a MCP server that lets AI agents search sessions. It is a managed service, so if recordings must stay on your own servers, one of the self-hosted tools above is the better fit. If not, start on the free plan and see pricing when you grow.
Frequently asked questions
What is the best open-source session replay tool?
It depends on what you need around the replay. OpenReplay is the most complete self-hosted replay product for debugging. ClickStack fits teams that want replays inside an OpenTelemetry observability stack. Sentry suits teams already running Sentry for errors. Umami covers light replay needs on a privacy-first analytics tool. rrweb is the library almost all of them build on, if you want to build your own.
Is self-hosted session replay cheaper than SaaS?
Storage is cheap: even 100,000 sessions a month at a few hundred kilobytes each costs only a few dollars a month in object storage. The real cost is compute for ingest and search, plus engineering time for upgrades, backups, scaling and security. Below a few hundred thousand sessions a month, a managed service is usually cheaper once that time is counted. At high volume, or with strict data-residency rules, self-hosting can win.
Can I self-host PostHog session replay?
Only in a limited way. PostHog supports a single-server "hobby" Docker deployment that it recommends for small volumes (around 1,000 recordings a month), and it no longer supports new Kubernetes deployments. For production volumes PostHog recommends its cloud.
Does OpenTelemetry support session replay?
Not as an official signal yet. OpenTelemetry defines session.id and session.start/session.end in its semantic conventions (still in development status), and vendors use that ID to link replays to traces. In October 2026 ClickStack started reading the OpenTelemetry session.id for its replay views, and Elastic and Grafana are building replay on top of their OpenTelemetry-based browser SDKs.
Is open-source session replay GDPR compliant?
Self-hosting helps with data residency because recordings stay on your infrastructure, but it does not remove the need for a legal basis, consent where required, masking of personal data and a retention policy. Those obligations are the same for open-source and SaaS tools.
Session replay with errors and network data, without running the infrastructure. Try LogNroll for free.