Pay for what you index — not per seat.
Free to start. Usage-based from there. Enterprise self-host on request. No per-seat multipliers, no features hidden behind “Contact Sales.”
For tinkering and early-stage projects.
- →10,000 rows indexed
- →10,000 / month queries
- →Sync: Every 15 minutes
- →Live-tail streaming (capped — try it, don’t run prod on it)
- →Up to 3 projects, 3 teammates
- →Community support
For real products with real users.
- →500,000 rows indexed
- →100,000 / month queries
- →Sync: Every 5 minutes
- →Production-grade live-tail streaming
- →Webhook ingest, smart feeds, access rules
- →Up to 10 projects, 10 teammates
- →Email support
For growing engineering teams.
- →5,000,000 rows indexed
- →500,000 / month queries
- →Sync: Every minute
- →Custom transforms
- →Multiple environments
- →Up to 25 projects, 25 teammates
- →Priority email support
For internet-scale products.
- →25,000,000 rows indexed
- →Unlimited queries
- →Sync: Every minute
- →Everything in Team
- →Higher rate limits
- →Unlimited projects and teammates
- →Priority support
For regulated, self-hosted deployments.
- →Unlimited rows indexed
- →Unlimited queries
- →Sync: Every minute
- →Full self-host, licensed via signed JWT
- →SSO, audit logs, SOC 2 report
- →Dedicated support, SLA, custom contracts
- →Unlimited everything
- →Dedicated support
The fine print,
in plain words.
What counts, what happens at a limit, and how self-host works — answered straight.
What counts as a row?
Any record you’ve indexed into a lens. If your products table has 15,000 rows and you’ve declared a lens on it, that’s 15,000 rows. Duplicates and ghost rows don’t count — we dedupe at ingest.
What counts as a query?
A search, similarity, query, or feed call from your app. Admin actions in the Console don’t count. Failed requests don’t count.
What happens when I hit a tier limit?
Nothing dramatic, and definitely not a shutdown. Every plan — Free included — uses a warn → grace → gentle slowdown model so you’ve always got room to breathe. Your data is never deleted. Your app is never taken down.
80% used: a friendly email and dashboard nudge. Nothing changes for your users.
100% used: you slide into a 7-day grace window with up to 10% headroom. Everything keeps running — ingest just moves to a slower lane, and your searches, queries, and live feeds are untouched.
After grace: ingest takes a breather until your next cycle. Your live app keeps serving the whole time — search, similarity, queries, and open subscriptions all stay healthy. Upgrade whenever you’re ready and full speed returns instantly.
Hard resource caps (projects, team members, environments) are about scope, not cost, so they stop at the number you chose — grace doesn’t apply there.
What's the fair-use grace on every plan?
10% headroom for 7 days the moment you cross 100%. During grace, your users see no change — reads stay full speed, ingest just runs in a slower lane, and we label it clearly in your Console so nothing’s a surprise. After grace, ingest catches up automatically on your next billing cycle. No surprise bills. No deleted data. No overnight shutdowns. We’d rather keep your app alive and earn the upgrade than take you down to make a point.
Can I run SemiLayer in my own infrastructure?
Yes, two different ways. Runners are available on every plan, Free included — they put the data plane inside your VPC so your data never crosses the internet, while the Console stays hosted by us. No VPN, no inbound ports, just an outbound WebSocket. For teams whose compliance model requires the whole control plane on their own infra, Enterprise self-host ships the entire stack under a signed license. Runner handles “my data stays in my network”. Enterprise handles “everything runs on my servers”.
Can I cancel anytime?
Yes. Monthly cancels at end of cycle; yearly plans don’t auto-renew. Your data is yours — export at any time.
What currency?
All prices USD.
Ready to bolt on
intelligence?
Take it for a spin, ship in an afternoon, upgrade when it earns it.