Pricing for AI Tools
Usage-based vs seats, thoughts on pricing
Someone asked me yesterday: "How much will Brainz Lab cost?"
I froze.
I've spent weeks on architecture, days on design, hours on documentation. Pricing? I had a blank Google Doc with "TODO: figure out pricing" at the top.
Time to actually figure it out.
The problem with copying others
My first instinct was to look at competitors. Datadog charges by host and log volume. Sentry charges by events and seats. New Relic has this complicated "data plus users" thing.
They all have one thing in common: unpredictable bills.
I've seen Datadog invoices that made CTOs cry. "$47,000 this month. We left debug logging on in production." That's not a pricing model. That's a trap.
I don't want my users to be afraid of using the product.
The AI problem
Here's what makes this harder: AI tools don't fit traditional models.
Per-seat pricing assumes humans are the users. But what if Claude is querying your logs 24/7? Is that one seat? Zero seats? Infinite seats?
Usage-based pricing makes more sense for AI workloads. But it creates anxiety. Every MCP call has a cost. Every log has a cost. You start optimizing for the bill instead of for the product.
Neither feels right.
What I landed on
Tiers with predictable caps.
Free: $0/month
- 1GB logs
- 1000 errors
- 7 days retention
- Full MCP access
Pro: $49/month
- 50GB logs
- 10K errors
- 30 days retention
- Priority support
Enterprise: Custom
- Unlimited
- 1 year retention
- SLA, dedicated support
You know what you're paying. You know what you're getting. No surprises.
The free tier is generous on purpose. I want people to actually use it, hit the limits, and think "yeah, $49 is worth it."
The self-hosted twist
Here's what I'm most excited about: self-hosted users don't pay for usage.
Run it on your own servers? Unlimited everything. Your hardware, your costs. I don't meter it.
What do self-hosted users pay for? Support, enterprise features, managed upgrades, SLAs.
This feels right. Heavy users who self-host aren't punished for using the product heavily. They're rewarded with lower costs.
Light users who want convenience pay for the cloud. That's fair.
Digital Ocean partnership
Want to self-host but don't know where to start? We've partnered with Digital Ocean to make it easy.
We'll migrate you for free.
Spin up a droplet, point us to it, and we'll handle the setup. Your infrastructure, your data, your control—without the DevOps headache.
Why Digital Ocean? They get developers. Simple pricing, solid performance, no enterprise sales calls required. And here's the kicker: you'll save at least 35% compared to AWS, Azure, or GCP.
Same workloads. Same performance. A third less on the bill. That's real money back in your pocket.
This isn't a sponsorship. I just think it's the right fit for what we're building.
What I'm still figuring out
AI-specific features are tricky.
Anomaly detection, root cause analysis, predictive alerting—these require compute. Should they be included? Separate line items? Premium tier only?
I genuinely don't know yet. I'm going to start inclusive and see what happens. If AI features blow up my costs, I'll adjust.
The nice thing about being early: I can experiment. Nobody's locked into contracts. I can change things.
The uncomfortable truth
I'm not optimizing for revenue right now.
If I wanted to maximize revenue, I'd do what Datadog does: per-GB pricing, per-host pricing, consumption traps everywhere.
But I'm optimizing for adoption. Get developers using the tools. Build the habit. Prove the value.
Revenue follows adoption. Not the other way around.
Ask me about pricing again in a year. Hopefully I'll have actual data instead of spreadsheet guesses.
— Andres