Ruby is the language for AI.
At Coolhand Labs we help manage and improve production AI improvement loops across many clients, running on many types of codebases. As deep Rubyists, we try to stay objective and neutral about different implementations but, lately, we've noticed a trend:
Token use across our clients
We aren't going to stand on the sidelines and watch Ruby get, uh... sidelined in AI app development.
The challenge
So, when Coolhand Labs was selected as one of the year's notable new Ruby startups at the 2026 RubyConf, our founder used the time on the main stage to make a commitment to the Ruby community:
"Let's make Ruby the best language for building AI apps, just like it's the best language for building everything else."— Michael Carroll, Founder of Coolhand Labs
View our RubyConf presentation
Curious how Coolhand stacks up on Ruby support specifically? See RubyLLM + Coolhand.
The patterns and capabilities of AI product development are far from set in stone.
Good open source tools require creative ideas, experimentation, ownership, and a lot of love…
No language community has more of that than Ruby.
Coolhand Labs is making two big* commitments to the Ruby AI ecosystem:
* big for a bootstrapped startup like us 😅
We are making Coolhand 100% free for Ruby apps
Free until 2027 — and maybe further, if that's what it takes to make Ruby the top language for AI apps.
What does that mean in practice? We won't count tokens coming from Ruby apps against your usage limits.
That's basically giving your Ruby app no limits on our free tier. (JS and other languages still need to
pay, though. 😉)
Until the end of the year, Ruby apps get Coolhand 100% free.
Free
-
Up to 10M tokens/week
Tokens from Ruby apps don't count against this limit! - Cost insights dashboard
-
5 optimization suggestions/week
Unlimited for Ruby apps -
Direct code fixes (PRs) at $2 each
Free for Ruby codebases! - 21-day data retention
Pro
-
Up to 100M tokens/week
Tokens from Ruby apps don't count against this limit! - Cost & performance insights
- No optimization cap
-
10 PRs/week included
Unlimited PRs for Ruby apps - Additional PRs at $1 each
- 90-day data retention
Power User
-
No token cap
Ruby tokens aren't charged! - Cost & performance insights
- No optimization cap
- Unlimited PRs
- HIPAA compliance
- Volume pricing available for annual plans
Free
-
Up to 10M tokens/week
Tokens from Ruby apps don't count against this limit! - Cost insights dashboard
-
5 optimization suggestions/week
Unlimited for Ruby apps -
Direct code fixes (PRs) at $2 each
Free for Ruby codebases! - 21-day data retention
Pro
-
Up to 100M tokens/week
Tokens from Ruby apps don't count against this limit! - Cost & performance insights
- No optimization cap
-
10 PRs/week included
Unlimited PRs for Ruby apps - Additional PRs at $1 each
- 90-day data retention
Power User
-
No token cap
Ruby tokens aren't charged! - Cost & performance insights
- No optimization cap
- Unlimited PRs
- HIPAA compliance
- Volume pricing available for annual plans
We're funding the Ruby AI community directly
Even though we're a bootstrapped startup, we want to put some of the money from our initial traction back
into the community.
So we're funding 5 rising open source Ruby + AI
projects with a no-strings-attached $500 donation in
2026. We define "rising" as having between 50 and 500 GitHub stars. We'll consider submissions on a
rolling basis, until all the funding is spent.
Nominate an open source Ruby project (or submit your own)
New Ruby tools we'd love to see (and help fund!)
Idea #1
Batch inference tooling
ruby_llm just landed a cross-provider batch API on main — help harden it with smarter provider routing, retries, and cost tracking.
Idea #2
Composable, DRY skills
No agent-skill framework lets you share instructions across skills without copy-paste — build the missing includes/templating layer.
Idea #3
Skill linting
Existing linters check frontmatter and naming — none catch duplicated prompt content or real cross-platform compatibility.
Idea #4
Incremental skill evals
Today's eval frameworks re-test a whole skill every run — build tooling that scopes evals to just the parts a diff actually touched.
Idea #5
Rails-native agent harness
No Rails-native framework exists for building/running AI agents — build a harness with agent loops, tool-calling, and memory that fits Rails conventions (ActiveJob, ActionCable, etc).