Skip to content
Tridot
MENU

SOLUTIONS / CONVERSATION ANALYTICS

Catch risk in the turn it appears.

Send a conversation turn and get a risk assessment back in real time: a severity, a suggested reaction and a score per category, from self-harm and minors’ safety to jailbreak attempts. The same turn events also feed reports on drop-off by turn and churn signals.

API · Sandbox
Create developer account Try it in the Playground Discuss this service
Availability
API · Sandbox
Interfaces
REST · MCP

WHAT IT DOES

Real-time, per-turn risk detection for character conversations.

What it detects
  • Self-harm
  • Minors’ safety
  • Sexual content
  • Violence
  • Harassment
  • Personal information
  • Dependency
  • Jailbreak attempts
  • Illegal activity

Real-time assessment

Every turn returns a severity from none to critical and a suggested reaction: monitor, intervene, block or escalate to a person.

Conversation-level state

Each conversation keeps a rolling risk level, so a slow escalation across several turns is visible, not only a single bad message.

Reports

Drop-off by turn, churn signals and risk findings by category for any date range.

Built for deletion

Reasons never quote the user’s text, and one call deletes everything stored for a conversation.

Why it matters.

Conversation logs tell you something went wrong after the fact. A safety team needs to know in the turn it happens, and which conversations to look at first.

INPUT / OUTPUT

From input to output.

INPUT

A conversation turn · up to 10 previous turns · pseudonymous user id

PROCESS

Score the turn per risk category and track the conversation’s level over time

OUTPUT

Severity · suggested reaction · category scores · reports

Call it over REST or MCP with one API key. Sandbox keys are never billed; the full request and response reference, with the OpenAPI specification, will be in the developer portal docs when the portal opens (join the waitlist for now).

MORE SERVICES

Works well with.

Character API

Chat with a character that remembers, stays in character and stays safe.

API · Sandbox

Guardrail

Safety memory for AI characters.

Preview · Sandbox

LET’S BUILD WHAT’S NEXT

The next chapter
of human–AI interaction.

For technology partnerships, product integrations, and investment conversations.

Talk to Tridot