Klavis AI
Provides long-horizon coding and agentic tool-use datasets with programmatic verification, granular rewards, and Dockerized environments for RL and SFT.
A platform that provides high-quality coding and agentic tool-use data for training AI models. It solves the problem of scarce, low-quality training data for long-horizon tasks by offering datasets with programmatic verification, granular rewards, and Dockerized environments. It sells to AI research labs, enterprises, and developers building frontier models, targeting technical teams in mid-market and enterprise companies. It differentiates by delivering realistic, state-mutating workflows across 600+ real tools and SaaS apps, with verifiable rewards, unlike demo-only or static alternatives.
Key features
- Long-horizon coding tasks
- Programmatic verification
- Granular rewards
- Dockerized environments
- Code, test, and debug loops
- 600+ real tools and SaaS apps
- Live SaaS apps and production MCP servers
- Deterministic, rubric, and LLM-judge rewards
- X<0.1/day
- Blog
- Referral program
- Community
- API
- Docs
- Engineering teams
- Data analytics teams
- Enterprises
Recent go-to-market activity
A sample of the latest public posts. The full timeline, every channel and change alerts are available with a free account. See the full timeline →