Back to directory
Cloud & DevOpsnicolasmontone/jev-evals
jev-evals
Rubric-based eval harness for LLM/agent outputs, backed by the typesafe-ai/jev evaluation model via Vercel AI Gateway. Cheap enough (~$0.04/1M input tokens, one round trip per case) to run on every PR.
Suggested install command
npm view jev-evalsAlways inspect the linked repository and skill instructions before running commands. Skills are instructions; permissions and execution still matter.
Compatibility
Agent support matrix
3 supported
| Agent | Status |
|---|---|
| Claude Code | Supported |
| OpenCode | Not listed |
| Cursor | Supported |
| MCP | Not listed |
| GitHub Copilot | Not listed |
| Windsurf |