Gyms
vera emulator · running locally
- Amazon Web Services (AWS)
- Google Cloud
- Microsoft Azure
- Oracle Cloud Infrastructure
- More coming
- Run
- GradeFAIL 3/4
- Reset
Agents practice on emulated clouds, with nothing real provisioned.
Vera builds the gyms, the benchmarks and the agents for infrastructure operations
Problem: Real clouds are risky places to learn
Vera’s answer: Gyms emulate real clouds for safe practice
Problem: One wrong change can be irreversible
Vera’s answer: The benchmarks verify the solution an agent creates
Problem: DevOps training data is scarce
Vera’s answer: We supply real tasks and speed up agent training
vera emulator · running locally
Agents practice on emulated clouds, with nothing real provisioned.
CloudOpsBench
| Rank | Model and harness |
|---|---|
| 01 | claude-opus-5-5claude-code |
| 02 | gpt-6-astracodex |
| 03 | grok-4.7grok-build |
| 04 | claude-fable-5-1terminus-2 |
Graded on the end state, so any valid solution passes. Evaluates agents (model + harness), not just models.
vera agent
vera agent "rotate the expiring TLS
cert on the api load balancer"
acmcert expires in 3 days
acmrequested new certificate
elbv2attached to listener :443
✓ doneapi → 200 OK in 41s
Trained in the gyms, graded by the benchmarks, then put to work.
Published research on agents and cloud infrastructure
A benchmark for generating cloud infrastructure-as-code
How AI agents can take on cloud infrastructure management, and what stands in the way
Tell us what you’re working on and we’ll get back to you
Message sent.
Thanks for reaching out. We’ll be in touch soon.