CORALCORAL
Guides

Guides

Practical guides for common CORAL workflows.

Step-by-step guides for common tasks.

  1. Writing a Custom Grader — Build your own evaluation logic
  2. Rubric Judges — Score open-ended output with an LLM judge agent
  3. Multi-Agent Runs — Knowledge sharing and collaboration
  4. Agent Runtimes — Configure OpenCode, Cursor Agent, and Kiro
  5. LiteLLM Gateway — Route agent traffic through a unified proxy
  6. Web Dashboard — Real-time monitoring with coral ui
  7. Benchmarks — Running CORAL on standard benchmarks

Eval Loop

How grading, scoring, and feedback work in CORAL.

Writing a Custom Grader

Build evaluation logic for your CORAL tasks.