The honest arithmetic on replacing Loom with Cap

Cap is the only project that genuinely replaces Loom, and for a small team it is still the more expensive choice. Loom Business costs $180 per seat per year. Our self-host floor is $600/yr for the server alone — before you store a single recording. At one seat, self-hosting costs about $420 more per year; at three seats, about $60 more. It only pulls ahead near ten seats, where Loom bills $1,800.

That conclusion is the opposite of what most "open-source Loom alternative" lists tell you, and the reason is always the same: they compare a subscription against a server rental and stop there. Video does not work that way. Every recording you keep occupies storage that only grows, and every viewer who presses play pulls that file back out across the network. The bill tracks how many people watch, which is the one number a VPS price cannot tell you.

Who should not read this article: if you have fewer than about eight people who need shareable video, self-hosting will cost you more than Loom and give you a worse viewing experience. Buy the subscription, or use Cap's own cloud, which is cheaper than Loom.

1. What Loom actually costs

Loom Business is $180 per seat per year on annual billing, verified from Loom's own pricing payload. The seat ladder is linear, which makes the comparison unusually clean:

Two caveats that cut in opposite directions, both worth stating plainly. Loom's free tier caps recordings at five minutes, which is the wall most people hit rather than a seat limit. And the $600 self-host figure is a floor: it is a 8 GB VPS plus a modest operations allowance, and it contains no storage, no egress, and none of your time. The real self-hosted number is higher than the table shows, which makes the small-team verdict stronger, not weaker.

2. The line every other comparison leaves out

Cap requires S3-compatible object storage. That is not an optional extra you can defer: the deployment includes MinIO precisely because recordings have to live somewhere. Once storage is in the picture, so is egress, and egress is where video billing gets unpleasant.

The pricing model you choose matters more than the per-gigabyte rate:

R2's free egress is the reason it wins for this workload. A self-hosted video instance is egress-heavy by definition — the whole point is that colleagues watch the recordings — so a provider that charges nothing for people watching removes your least predictable line item entirely.

We list Vultr as NOT_FOUND rather than filling in a plausible number. Their object storage page returns an error to automated requests, and we would rather publish a gap than an invention.

There is a second variable cost people miss: Cap transcodes video on your server. The media server image installs FFmpeg, so CPU load scales with how much people upload, not just how much they watch. Cap publishes no minimum hardware specification, so our 8 GB sizing is our own judgement based on running 5 services plus transcoding, and we label it as such rather than presenting it as a vendor requirement.

3. The licence has two halves, and they are usually reported wrong

Cap's licensing trips up almost every article that covers it, because there are two separate things happening.

The code is a composite: the root LICENSE file is AGPLv3 for the bulk of the project, while licenses/LICENSE-MIT covers the cap-camera* and scap-* crate families. Because GitHub's API cannot resolve a split like that, it reports NOASSERTION — which is why you will see Cap listed as having "no licence." It has two.

The binaries Cap distributes are a different question. Using Cap's official desktop build for commercial purposes requires a $29/yr Desktop License. Cap's documentation is explicit that this licence "does not apply to versions of the app you build yourself from our source code, only to binaries that we distribute." So: compile it yourself and the AGPLv3 terms are all that apply; download their build and use it at work, and you owe $29/yr.

Neither half changes the arithmetic above much — $29/yr is small next to a $600 server — but conflating them into "Cap costs $29" or "Cap is fully free" both mislead.

4. Change the default secrets before you expose it

This is Cap's own documented warning, not a hypothetical we constructed. The docker-compose.yml in the repository ships with placeholder secret values, and because they are committed to a public repo, everyone knows them. Cap's self-hosting docs state that leaving them in place lets an attacker:

• forge authentication sessions, via NEXTAUTH_SECRET
• decrypt sensitive database fields, via DATABASE_ENCRYPTION_KEY
• spoof webhook requests

Generate three fresh values with openssl rand -hex 32 before the instance is reachable from the internet. One more operational detail from the same docs: login links are printed to your server logs until you configure a Resend API key, so anyone with log access can sign in as anyone.

We flag this because "it deploys with Docker Compose" is often presented as though it means "it is safe by default." Here it does not. Compare Docmost, which refuses to start at all until you set a real secret — a better default, and a useful reminder that convenience and safety are not the same property.

5. What you give up

Cap replaces Loom's core loop honestly: record, upload, share a link. That is more than Screenity or OBS can claim. But three things Loom sells are not things a single server gives you.

CDN delivery. Loom serves video from a global edge network. Your self-hosted instance serves it from one region. Colleagues on another continent will notice, and no amount of tuning fixes the distance.

Transcription and viewer analytics. These are part of what Loom's per-seat price buys. Do not assume parity here.

Someone else's on-call rotation. You now own upgrades, the storage bill, and the security posture described in section 4.

If none of that appeals but you still want out of Loom's pricing, the useful middle path is Cap's own cloud at $12/user/mo monthly or $8.16/user/mo annual — cheaper than Loom Business, with no server to run. We mention it because our job is the cheapest sensible exit, not the most self-hosted one.

6. Verdict

• 1 to 3 seats: stay on Loom, or move to Cap's cloud. Self-hosting costs $420 to $60 more per year before storage, and gives you worse playback.
• Around 8 to 10 seats: self-hosting starts to make sense. At ten seats Loom is $1,800 against a $600 floor, a $1,200 gap wide enough to absorb storage, egress and some of your time.
• Whatever the size, budget the storage line. Use Cloudflare R2 for free egress, and remember the cost scales with viewers rather than recorders.
• Do not deploy the default compose file as-is. Three openssl rand -hex 32 values first.

Cap deserves the attention it gets. It is the most actively developed project we looked at in this category, with a real release cadence and no known CVEs at the time we checked. Our disagreement is not with the software; it is with the claim that self-hosting it saves a small team money. For most teams reading this, it does not.