ABOUT SENTIQRA
We started with Trust & Safety. We built decision evidence infrastructure.
Sentiqra began because moderation teams needed more than a toxicity score - they needed to know what was detected, how strongly it matched, and whether the same case could be reproduced later.
The boundary is foundational: Sentiqra produces decision evidence; your policy makes the decision.
That same need - reproducible, auditable analysis - exists wherever automated decisions affect users, transactions, or investigations. All 12 production analyzers expose the same deterministic product contract and versioning guarantees to serve it.
12
Production analyzers
13
Catalog products
43
Supported languages
Why we built Sentiqra
Sentiqra began with Trust & Safety.
Moderation teams needed more than a single toxicity score. They needed to know what was detected, how strongly it matched, and whether the same case could be reproduced later.
We found the same problem elsewhere.
Security teams need repeatable threat analyses.
Crisis analysts need structured multilingual signals.
Finance teams need versioned event evidence.
Information-integrity teams need manipulation signals without outsourcing political judgment to a generative model.
So all 12 production analyzers expose the same deterministic product contract:
Different taxonomies. The same architectural principle:
Same normalized input + same pinned detection configuration = same analysis result.
What we believe
Reproducibility over guesswork
If a classification matters today, teams should be able to examine and replay it later. Sentiqra versions the components that define classification behavior rather than relying on free-form generative output.
Evidence over assertions
A signal score is only useful when you know what it represents. We document our taxonomies, publish benchmark methodology, and separate reproducibility claims from accuracy claims.
View benchmarks →Privacy through architecture
Production customer inputs are not used to train Sentiqra's shared detection models. Our data-handling documentation explains what is processed, what is retained, and what is not.
Read our data practices →Humans own consequential decisions
Sentiqra provides classification evidence. Your policies, applications, analysts, and reviewers determine what happens next. We do not believe a signal score should silently become an enforcement, trading, security, or political judgment.
Built to be inspected.
You don't need to trust an About page to evaluate Sentiqra.
Verify reproducibility
Run the same pinned analysis repeatedly and compare the canonical result.
Verify it yourself →
Review benchmark performance
See benchmark methodology, datasets, versions, and per-language Top-3 Hit Rate results.
View benchmark performance →
Review the architecture
Understand the model path, versioning, retention, and training-data boundaries.
How Sentiqra works →
Review security and governance
See current technical, privacy, contractual, and compliance documentation.
Visit Trust Center →
Our journey
2024
Research, taxonomy & first API
Developed the first Trust & Safety taxonomy - 200 indicators across 20 policy groups - and shipped API v1.0 with the two primitives that still define the platform: pinned detection configurations and signed response receipts.
2025
Multilingual robustness
Extended detection across languages with per-language evaluation and the Top-3 benchmark methodology we still publish today.
2026
Multi-domain platform
Expanded the deterministic product contract to a 13-analyzer catalog - 12 generally available: safety, security, discrimination, emotion, crisis, market, disinformation, customer-service, dating-safety, fraud, insurance, and wellness intelligence; plus restricted Information Warfare. Published the benchmark catalog, opened the public verification playground, and added plan tiers with human service commitments.
Who we are
Sentiqra is built by a small engineering team headquartered in the European Union, working on content moderation, security, and analysis systems.
Engineering-led
The team maintains the taxonomies, analysis pipelines, and production infrastructure directly - the people who answer your technical questions are the people who built the system.
Evidence, not assertions
Every analysis is pinned to an immutable configuration and every successful result carries a signed receipt. If we can't show it to you, we don't claim it.
European by design
Headquartered in the EU. Infrastructure locations, subprocessor relationships, and transfer basis are disclosed in the Trust Center; an EU-residency option is on the Trust Center roadmap. GDPR obligations are handled under a self-assessed posture backed by documented controls and retention schedules, not just stated in policy documents.
Reachable
Technical questions from paid plans are answered by an engineer within 1 business day - no ticket queue. Free-plan developers are covered by the docs and community; higher plans add proactive and dedicated support. Get in touch.
Working with us
Service is part of the product, not an upsell. These are the commitments attached to each plan.
Onboarding call + quarterly office hours. A real walkthrough of configuration pinning, threshold selection, and receipt verification - not a docs link.
Named customer engineer + Success Package. Kickoff call, threshold workshop, migration review, and 30/60/90-day check-ins. The person who onboards you stays your contact.
Customer success manager + technical account manager. Proactive reviews of usage, configurations, and upcoming analyzer changes before they reach your traffic.
Dedicated implementation team. Deployment architecture review, annual configuration audit, and a direct line to the people who run the platform.
Full plan details on the pricing page.
Build on evidence you can replay.
Start with 5,000 analyzer runs per month across all 12 production analyzers.
No card required.