Evidence, not answers.
A text-analysis API with reproducible signals and verifiable receipts.
Analyze text for documented safety, security, fraud, and customer-workflow signals. Keep control of the policy decision, and reproduce analyses using supported pinned configurations.
Talk to Solutions Engineering · 13 analyzers · 43 languages platform-wide · 8–41 per analyzer · 5,000 free runs/month
Click an indicator or highlighted segment to trace it.
Every run returns a configuration ID, fingerprint, caller-keyed input commitment, and signature you can verify later. Submitted text is not stored.
BUILT FOR TEAMS WHO CAN'T AFFORD TO BE WRONG
13
Analyzer Catalog
12
Generally Available
43
Supported Languages
Versioned
Reproducible Results
Non-generative semantic analysis · Submitted content is not retained · No training on customer inputs
SIGNALS TO BUSINESS RESPONSES
One message. More than one business response.
A message can require more than one business response. Sentiqra returns ranked, qualifying signals rather than forcing every message into a single winning label. It separates detection from customer policy, allowing your systems to route, escalate, or review relevant secondary signals.
By default, analysis reports up to three qualifying signal groups per analyzer, with finer-grained child labels where available. It can report fewer groups or none. Reduced output profiles may show fewer signals.
One message → Qualifying signals → Your policy
Possible customer-controlled workflows, not actions executed by Sentiqra.
Route beyond the main topic
A secondary concern can have a different owner. Your router can assign a primary team while preserving relevant signals for specialist review.
Prioritize using combinations
Your policy can consider qualifying signals together instead of treating the winning label as the only concern. Measure the effect on unnecessary alerts and workload.
Preserve context for reviewers
Give reviewers the returned indicators, not just one winner. Signals describe matches in the text; they do not establish intent or prove an allegation.
THE PROBLEM
Automated decisions get challenged. Your evidence has to hold up.
A user appeals a flag. A regulator asks why an account was frozen. An analyst wants to know why the model missed something last Tuesday. If your text-analysis stack is a black box - or worse, a generative model whose output drifts between calls - you can't answer any of these questions with a straight face.
Generative classifier
- - Output varies between calls
- - Prompt- and model-dependent
- - Historical decisions hard to replay
- - One opaque score or paragraph
Sentiqra
- Same input + pinned config → same output
- Pinned to an immutable configuration ID
- Past decisions reproducible while their configuration ID remains supported
- Structured, ranked, machine-readable signals
WHAT REPRODUCIBILITY UNLOCKS
Not a technical property. A business capability.
Defend every decision
- • Replay the exact analysis a user is appealing
- • Show auditors which config produced which result
- • Preserve evidence behind automated actions - Sentiqra stores no content; your system retains the original input plus the verification receipt needed for future replay
Ship configuration changes without fear
- • Diff old and new configs before you deploy
- • Run regression sets to catch behavior changes early
- • Keep production on a pinned config while you validate the next one
Investigate while the configuration is supported
- • Reproduce decisions from a stored config ID while that configuration remains supported
- • Separate analyzer behavior from policy logic when debugging
- • Know exactly when analyzer behavior changed, and why
ONE API. 12 ANALYZERS IN PRODUCTION.
Same deterministic product contract. Pick the signals your policy needs.
Safety & Trust
-
Trust & Safety
Harmful language, harassment, abuse
-
Discrimination Analysis
Protected-class and exclusionary language
-
Dating Safety
Romance scams, coercion, unsafe contact
Security & Fraud
-
Cybersecurity
Phishing, credential abuse, threat signals
-
Fraud & Scam Detection
Payment diversion, impersonation, lures
Intelligence & Risk
-
Insurance Intelligence
Claims, coverage disputes, fraud indicators
-
Crisis Intelligence
Conflict, displacement, humanitarian signals
-
Disinformation Analysis
Coordinated influence, manipulation techniques
-
Market Intelligence
Market-moving events, corporate actions, credit risk
Customer Intelligence
-
Emotion Analysis
20 distinct emotions, ranked
-
Customer Service Intelligence
Intent, urgency, escalation, routing
-
Wellness Intelligence
Wellbeing and human-state signals
Restricted - deployed under custom agreements: Info Warfare - Request an access briefing
PROOF, NOT PROMISES
Send the same text twice. Get the same signals twice.
Every successful analysis is pinned to an immutable configuration ID and returns a signed receipt. Store it. Replay it. Verify it later.
{
"signals": [
{
"analyzer": "trust-safety",
"code": "insults_derogation",
"label": "Insults & Derogation",
"score": 0.9,
"level": "category",
"confidence": "very_high"
},
{
"analyzer": "trust-safety",
"code": "insult",
"label": "Insult",
"score": 0.9,
"level": "label",
"parent": "insults_derogation",
"confidence": "very_high"
},
{
"analyzer": "trust-safety",
"code": "belittling",
"label": "Belittling",
"score": 0.9,
"level": "label",
"parent": "insults_derogation",
"confidence": "very_high"
}
],
"verification": {
"version": 3,
"configuration": "sqcfg_ZZW77NM9",
"fingerprint": "fp2_8a72c4...19f",
"input_commitment": "ic3_31bd5e...a42",
"signature": "sig_ed25519_4c1f...e2a",
"key_id": "edk_af4f218e",
"binding_key_id": "bk_31bd5e42f7a019c8"
}
}
{
"signals": [
{
"analyzer": "trust-safety",
"code": "insults_derogation",
"label": "Insults & Derogation",
"score": 0.9,
"level": "category",
"confidence": "very_high"
},
{
"analyzer": "trust-safety",
"code": "insult",
"label": "Insult",
"score": 0.9,
"level": "label",
"parent": "insults_derogation",
"confidence": "very_high"
},
{
"analyzer": "trust-safety",
"code": "belittling",
"label": "Belittling",
"score": 0.9,
"level": "label",
"parent": "insults_derogation",
"confidence": "very_high"
}
],
"verification": {
"version": 3,
"configuration": "sqcfg_ZZW77NM9",
"fingerprint": "fp2_8a72c4...19f",
"input_commitment": "ic3_31bd5e...a42",
"signature": "sig_ed25519_4c1f...e2a",
"key_id": "edk_af4f218e",
"binding_key_id": "bk_31bd5e42f7a019c8"
}
}
Immutable config
Configs are frozen. Fixes create new versions; old ones never change.
Signed receipt
Every successful analysis carries a result fingerprint, input commitment, and Ed25519 signature you can verify offline.
Historical replay
Reproduce a past analysis by resubmitting input against its config ID, while that configuration remains supported under the published support policy or your contract.
BESIDE YOUR LLM, NOT AGAINST IT
Use generative AI for what should be flexible. Use Sentiqra for what must not drift.
Use an LLM for
- • Drafting a moderator explanation
- • Summarizing a case
- • Translating policy to natural language
- • Synthesizing context from sources
Use Sentiqra for
- • Assigning a documented policy signal
- • Producing reproducible signal-match scores
- • Replaying a historical analysis
- • Feeding deterministic policy logic
Because Sentiqra is deterministic, identical inputs can be served from cache with the same receipt - the cached response is provably the same analysis, not merely a stored answer.
REST or MCP tool
NEXT STEPS
Three ways to check the claims.
See an output
Run sample text through the Playground and inspect the signals, fingerprint, and receipt it returns.
Try sample analysis →Check the limits
Browse the analyzer catalog - generally available vs restricted - with per-analyzer language coverage and input limits.
Browse the analyzer catalog →Make a call
Follow the quickstart for a first request, then the API reference for schemas, retries, and verification.
Read the API quickstart →Sentiqra isn't for every workflow. Here's when it fits.
Strong fit if you:
- • Need structured signals behind automated decisions
- • Operate multilingual safety, security, intelligence, or risk workflows
- • Have to reproduce historical decisions for audits or appeals
- • Need to defend automated actions to regulators or users
Probably not a fit if you:
- • Only need approximate general-purpose moderation
- • Have no audit, appeal, or reproducibility requirement
- • A free classification API is enough for your volume
- • Need image, video, or audio analysis - Sentiqra is text only
- • Want the API to make the moderation decision for you - it returns evidence; your policy decides
See it prove itself.
Don't take this page's word for it. Send a request, repeat it, compare.
5,000 analyzer runs/month · No card required · Talk to Solutions Engineering