Support quality
Check whether the assistant addressed the request, explained next steps or made an unsupported promise.
RESPAN / BEHAVIOR EVALUATION
Span-01 Lite: measure conversation behavior.
Evaluate conversation behavior with Respan’s lightweight model. Span-01 Lite has free upstream inference; this service offers a free signed-in workbench and charges one credit per successful API request.
Both workbenches are free after sign-in, even at zero balance. Only API calls use credits.
Answer type
Noul
A probability for each behavior
API credits
1 credit
per successful API request
Playground charge
0 credits
Sign in and run free
Span-01
Higher-accuracy tier. API: 200 credits per million input tokens, rounded up, with a minimum of 1 per successful request.
Span-01 Lite
Lightweight tier. Upstream input and output are free. Our API charges a minimum service fee of 1 credit per successful request.
Agent guardrails
Define the behavior you want to detect, such as sharing sensitive information or promising an unauthorized action.
Conversation monitoring
Evaluate several independent behaviors on the same conversation and inspect each probability.
Behavior evaluation workbench
Add up to eight yes/no behaviors. Use plain text or the conversation editor. Each behavior returns its own probability; no labels or score levels are needed.
Span · behavior evaluation
Free after sign-in · API uses credits
Upstream inference is free. Our API charges 1 credit per successful request; playground runs cost 0.
Draft saved in this browser · Up to 8 behaviors · 32 KiB per request · web runs are free.
Which behaviors appear in this response?
Run the example or add your own behaviors. Each result will show the model’s estimated probability that the behavior is present.
Web runs and saved configurations are private to your account. API history stores usage only.
UsageAPI pricing
OpenRouter lists free input and output. Our API charges 1 credit per successful request for the service. Signed-in web runs cost 0 credits; failed provider calls are not billed.
max(1, ceil(input_tokens × 0 / 1,000,000))
Upstream pricing and text/conversation requests verified on September 26, 2026. Output tokens are free. API calls need a positive balance; failed provider calls are not billed.
OpenRouter · Span-01 LiteOne endpoint, model-specific validation
Use a site API key with /v1/systemone and select respan/span-01 or respan/span-01-lite. Questions must use type: noul. The Code tab generates JavaScript, Python or cURL from your current input. Answers are at data.result.answers.
This service accepts up to 8 behaviors and 32 KiB per request. Conversation input is an input message array plus an assistant output message; arbitrary objects and bare arrays are not accepted. Web runs share the one-attempt-per-3-seconds limit.