CAP.01
RESPONSE EVALUATION
Evaluate AI outputs
Submit prompts, generated responses, references, and evaluation criteria for structured quality assessment.
API PLATFORM
Evaluate. Anywhere.
Connect EvalLoop AI with the systems you already use. Send evaluation inputs from your applications, workflows, or AI infrastructure and bring the results back into your own environment.
API ACCESS · CURRENTLY IN DEVELOPMENT
EVALLOOP API
DEVELOPMENT
// evaluation request
POST
/evaluate
{
"input"
:
"AI response to evaluate"
,
"reference"
:
"Expected behavior"
,
"criteria"
:
[
"factuality"
,
"relevance"
,
"instruction_following"
]
}
// evidence-backed evaluation
→
score + evidence + verdict
API.01 / CAPABILITIES
Connect your AI. Evaluate it.
The EvalLoop API is being designed as a programmatic layer for evaluation. Instead of manually moving responses into a dashboard, applications will be able to send evaluation inputs directly into EvalLoop workflows.
CAP.02
COMPARISON
Compare AI responses
Evaluate two or more candidate responses against the same criteria and identify meaningful quality differences.
CAP.03
RAG EVALUATION
Inspect grounded answers
Provide retrieved context alongside an AI response to evaluate grounding, relevance, completeness, and unsupported claims.
CAP.04
SAFETY
Test AI behavior
Evaluate responses against safety policies, prohibited behaviors, and custom evaluation criteria.
CAP.05
MULTIMODAL
Evaluate richer inputs
The platform is being designed to support evaluation workflows involving text, images, and other AI-generated outputs.
CAP.06
EVIDENCE
Results with reasoning
Evaluation results are designed to connect scores with criteria, evidence, and clear verdicts instead of returning unexplained numbers.
API.02 / WORKFLOW
From request to evidence.
API evaluation is designed to fit into an existing AI workflow without requiring the application itself to move into EvalLoop.
01
Send
Provide the prompt, response, context, reference, or other evaluation inputs your workflow already produces.
02
Evaluate
Apply the evaluation criteria appropriate for the selected workflow.
03
Inspect
Examine scores, evidence, criteria, findings, and evaluation verdicts.
04
Improve
Use evaluation findings to identify failures and improve the underlying AI workflow.
API.03 / DESIGN
Built around evaluation.
EvalLoop API is intended to make evaluation a usable part of an AI development workflow rather than a separate manual step.
PRINCIPLE.01
Your AI stays with you
Your application or AI system remains under your control. EvalLoop receives the inputs that your workflow intentionally sends for evaluation.
PRINCIPLE.02
Evaluation should be explainable
Results should connect scores to criteria, evidence, and verdicts so teams can understand why an output passed or failed.
PRINCIPLE.03
Fit existing workflows
Evaluation should work alongside existing AI applications, RAG systems, model pipelines, and testing workflows.
PRINCIPLE.04
Start simple
A useful evaluation should not require a large infrastructure change. The API is being designed around straightforward evaluation inputs and structured results.
API ACCESS — COMING SOON
Endpoint contracts, authentication, request schemas, response schemas, rate limits, and production availability will be documented before public API access.