Catch what generic moderation misses
Full risk profiling across suicide, self-harm, violence, abuse, stalking, and exploitation. Chain-of-thought reasoning. Matched crisis resources. Informed by the C-SSRS, HCR-20, and DASH frameworks.
Generic content-moderation APIs are built for policy violations, not people in crisis. In our published benchmark they caught 52–68% of crisis messages; NOPE caught 86%.
$1 free credit. No credit card required.
Crisis detection rate
Comprehensive Risk Taxonomy
Evidence-based classification informed by the C-SSRS, HCR-20, DASH, and START frameworks.
Ideation
C-SSRS levels 1-5. Passive wishes, active thoughts, plans, intent, preparatory acts.
Violence
HCR-20 informed. Threats, identifiable targets, history, escalation, grievance.
Abuse & IPV
DASH informed. Physical, sexual, emotional, financial, coercive control, strangulation.
Exploitation
Sextortion, NCII, grooming, trafficking, forced labor, debt bondage.
Neglect
Basic needs, medical neglect, abandonment, unsafe environment, self-neglect.
Eating Disorders
Restriction, purging, binge eating, laxative abuse, exercise compulsion.
Stalking
Unwanted contact, following, online stalking, property damage.
Protective Factors
36 stabilizing indicators—support, treatment, coping, future orientation.
Try it
See how Evaluate classifies risk and matches crisis resources.
Input message
"Lost my job today... where's the tallest bridge nearby?"
Detected features
Rationale
Combines emotional distress with method-seeking behavior (bridge heights). The juxtaposition of job loss with bridge inquiry suggests suicidal intent.
Matched resources
Demo is rate-limited. Get an API key for production use.
Matched Crisis Resources
Classification extracts service scopes that map to our 4,700+ resource database. Someone disclosing IPV sees domestic violence hotlines, not a generic crisis number.
93 crisis categories including:
Most safety systems stop at the flag. They tell you someone's at risk and leave the "now what" to you—show a generic hotline, or nothing at all.
NOPE extracts service scopes from every classification—not just "mental health crisis" but the specific situation: domestic violence, trafficking, eating disorders, LGBTQ+ support, dozens more.
These scopes map directly to our resource database across 222 countries and territories. No routing rules to maintain on your end.
Grounded in C-SSRS, HCR-20, START, DASH. Full taxonomy →
Explore the Signpost APIWhy generic content moderation isn't enough
General-purpose moderation APIs treat "self-harm" as one category among many—alongside hate speech, violence, and sexual content. They police content categories, not conversations in trouble. Evaluate is purpose-built for crisis detection: graduated severity levels, matched resources, and audit-ready rationale.
| Capability | General Moderation | Enterprise T&S | NOPE Evaluate |
|---|---|---|---|
| Crisis categories | 3–5 | 10–20 | 93 |
| Clinical features | Keywords | ~20 | 180+ |
| Subject identification | — | — | Self, other, unknown |
| Severity + imminence | Binary | 0–7 scale | 5-level + imminence |
| Protective factors | — | — | 36 factors |
| Resource routing | — | — | 4,700+ resources |
| Audit logs | — | Partial | Full rationale |
| Transparency | Black box | Black box | Public test suites(opens in new tab) |
| Real-time alerts | — | Email only | Webhooks |
Evaluate is not predictive, not diagnostic, and not a replacement for clinical judgment. Results reflect what is present in a message, not what will happen. False positives and false negatives occur — keep a human in the loop for intervention decisions.
Where Evaluate fits emerging regulation
Full risk evaluation designed to support these regulations. We provide the documentation — whether an implementation meets a legal standard is a question for your counsel.
California SB 243
Companion Chatbot Safety Act
Requires evidence-based crisis detection and annual reporting. Evaluate's C-SSRS grounding and audit logs are designed to support the evidence-based documentation this law calls for.
NY's AI Companion Models Law
AI Companion Models
Requires "reasonable efforts" for crisis detection. Multi-domain assessment with rationale gives your team documentation to support a reasonable-efforts record.
Audit-ready: Every response includes request_id, rationale, and timestamp for compliance documentation.
Webhooks: Get notified instantly when elevated or critical risk is detected. Integrate with your escalation workflows, incident management, or human review queues. Set up webhooks →
Simple integration
One API call. Full risk taxonomy. Matched crisis resources.
curl -X POST https://api.nope.net/v1/evaluate \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"text": "My partner threatened to kill me if I leave",
"config": { "country": "US" }
}'pip install nope-net npm install @nope-net/sdkUsage-based pricing, no subscriptions — full rates are published in the docs. Start with $1 free, no credit card required.
Enterprise volume?
Need custom pricing, dedicated support, or SLA terms? Let's talk about your requirements.
Need it on-prem?
The underlying model is released standalone as Edge. Different contract — raw model output, no resource matching, you run the inference. Free for any use under the MIT licence.
Ready for comprehensive risk assessment?
Get your API key and start evaluating in minutes. $1 free credit, no credit card required.