Guard Every API Call to AI
Enterprise AI workflows demand enterprise-grade security. NeuronEdge protects sensitive data across every Agentic AI call—before it reaches any LLM or SLM provider.
Private Beta · Request early access · Limited spots available
105+
Entity Types
Regex + NER ML detection
<1ms
Regex Latency
Pattern-based detection
5
Compliance Frameworks
HIPAA, PCI-DSS, SOC 2, FINRA, FedRAMP
0
Data Stored
Zero persistence
Built on Certified Infrastructure
SOC 2
Neon + Cloudflare
GDPR
Neon + Cloudflare
HIPAA
Neon BAA Certified
ISO 27001
Neon + Cloudflare
Core Detection Engine
Built for Enterprise AI Workflows
The foundation: a Rust-powered WASM engine that detects and redacts PII at the edge in under 1ms. Every capability below builds on this core.
Dual-Engine Protection
Regex + NER ML detection running in parallel across requests, responses, and tool call arguments. 102 pattern-based entity types at <1ms, plus contextual PERSON/ORG/GPE detection via Named Entity Recognition. Full bidirectional coverage.
Universal AI Gateway
One integration protects all providers. OpenAI, Anthropic, Azure, Bedrock, or any LLM-compatible endpoint. Agentic frameworks, RAG pipelines, chat apps—all covered.
Configurable Policies
Fine-tune detection per use case and per provider. Set provider-specific redaction rules, choose from HIPAA, PCI-DSS, FINRA, SOC 2, and FedRAMP compliance templates, or build custom policies from scratch.
Zero Data Persistence
PII is redacted in-memory, restored on response, never stored. Rust-powered WASM on Cloudflare edge. The architecture compliance teams love.
Platform
Beyond Detection. Complete AI Security.
From compliance automation to threat detection, NeuronEdge gives your security team the tools to protect, monitor, and govern every AI interaction.
Agent-Ready Security
AI agents make tool calls, chain reasoning, and process results autonomously. Every message shape is a PII vector — NeuronEdge covers them all.
Tool & Function Call Protection
PII in agent tool calls — arguments, results, parallel calls — redacted automatically in the streaming response.
Bidirectional Scanning
LLMs can hallucinate PII. Response-side detection catches generated SSNs, names, and addresses before they reach your application.
PII Extraction Defense
Detects adversarial prompts designed to trick models into revealing personal data. Log, warn, or block.
Compliance Automation
From audit-ready reports to one-click policy templates, NeuronEdge turns compliance from a quarterly scramble into a continuous guarantee.
Audit-Ready Reports
Generate PDF and JSON compliance reports with entity breakdowns, performance metrics, and zero-persistence attestation.
One-Click Compliance
HIPAA Safe Harbor, PCI-DSS, FINRA, SOC 2, and FedRAMP templates. Select a framework, deploy a policy.
Multi-Tenant Governance
Workspaces, policy inheritance, and org-level administration for complex organizations.
Operational Intelligence
Real-time visibility into what your AI is doing with sensitive data. Alerting, streaming analytics, and custom patterns for your domain.
Provider-Aware Policies
Different trust levels for different providers. Per-provider redaction rules, one API key.
Anomaly Alerts & Webhooks
Entity spikes, detection failures, high-density requests. Configurable alerts with webhook delivery.
Live Detection Stream
SSE endpoint for real-time entity detection events. Power SOC dashboards and SIEM integrations.
Custom Entity Patterns
Visual regex builder, test bench, and pattern versioning for domain-specific PII.
Simple Integration
One line. Every AI call protected.
Point your AI applications to NeuronEdge. We handle the rest—for OpenAI, Anthropic, and any LLM provider. Agentic frameworks, RAG pipelines, chat apps—all covered.
// Before: Direct OpenAI call
const openai = new OpenAI({ baseURL: "https://api.openai.com/v1" });
// After: Protected via NeuronEdge
const openai = new OpenAI({
baseURL: "https://api.neuronedge.ai/v1/openai",
apiKey: process.env.NEURONEDGE_API_KEY,
});
// That's it! PII is now redacted automatically.Pricing
Simple, transparent pricing
Start free and scale as you grow. All plans include our core PII protection engine.
Starter
Perfect for startups and small teams getting started with PII protection.
$199*/month
- 100,000 requests/month
- All 105+ entity types (Regex + NER)
- Tool call & function call redaction
- Token redaction format
- Compliance reports (JSON, General template)
- Email support
Professional
For growing teams that need advanced detection and compliance features.
$2,999*/month
- 5,000,000 requests/month
- Everything in Starter, plus:
- Response-side PII detection
- Provider-specific policy routing
- All compliance templates (HIPAA, PCI-DSS, SOC 2, FINRA, FedRAMP)
- PDF & JSON compliance reports
- PII anomaly alerts & webhooks
- Real-time analytics stream
- Prompt injection detection (log mode)
- Up to 5 workspaces
- 50 custom entity patterns with test bench
- Priority support
Enterprise
Dedicated infrastructure and support for enterprise compliance needs.
Custom
- Unlimited requests
- Everything in Professional, plus:
- Full prompt injection modes (log, warn, block)
- NER-powered response detection
- Unlimited workspaces & org hierarchy
- 200 custom patterns with versioning
- Scheduled compliance reports
- SSO & RBAC integration
- Dedicated support channel
- 24/7 priority support
- Dedicated success manager
* Prices shown are for the private beta period and are subject to change. Final pricing will be announced at general availability. All prices are in USD and exclude applicable taxes. Features and limits may vary.
Frequently asked questions
Can't find what you're looking for? Contact our team
- How does the dual Regex + NER detection engine work?
- NeuronEdge uses two detection methods in parallel: Regex (pattern-based) detects 102 structured entity types like SSN, credit cards, emails, and phone numbers at <1ms latency. NER (Named Entity Recognition) uses ML to detect contextual entities like PERSON, ORGANIZATION, and LOCATION names with ~500ms latency. Both run together for maximum coverage, or you can configure policies to use one or the other based on your speed vs. accuracy needs. The entire engine is built in Rust and compiled to WebAssembly—Rust's memory safety guarantees eliminate buffer overflows and use-after-free vulnerabilities common in C/C++, while its zero-cost abstractions deliver C-level performance without garbage collection pauses.
- What are the speed vs. accuracy trade-offs?
- Regex-only mode delivers <1ms detection latency—ideal for high-throughput applications where speed is critical. NER mode provides contextual accuracy for names and organizations that patterns can't reliably catch. The default "Both" mode (~500ms total) maximizes coverage. You can configure different policies per API key—use "Fast" for dev environments and "Accurate" for production healthcare data.
- Can I customize detection settings?
- Yes, extensively. The Policy Engine lets you configure detection methods (Regex/NER/Both), set NER confidence thresholds (0.80-0.99), enable specific entity types, and define custom patterns with a visual regex builder and test bench. Choose from pre-built compliance templates—HIPAA Safe Harbor, PCI-DSS, FINRA, SOC 2, and FedRAMP—or build custom policies. Set provider-specific overrides so different LLM providers get different redaction rules. Assign policies to API keys for multi-tenant applications or to separate dev/prod environments.
- What AI providers does NeuronEdge support?
- All of them. OpenAI, Anthropic Claude, Azure OpenAI, AWS Bedrock, Google Vertex AI, Mistral, and any OpenAI-compatible endpoint. One integration, universal protection.
- Does NeuronEdge work with Agentic AI frameworks?
- Yes. LangChain, AutoGPT, CrewAI, custom agents—NeuronEdge sits at the API layer, protecting all outbound AI requests regardless of framework or architecture.
- How does NeuronEdge protect data in AI API calls?
- NeuronEdge acts as a gateway proxy. It inspects every API call, detects and redacts 105+ PII entity types before requests reach AI providers, then restores original values in responses. Your sensitive data never leaves your control.
- What entity types can NeuronEdge detect?
- 105+ entity types across 8 categories: Identity (SSN, passport, driver's license), Financial (credit cards, IBAN, crypto wallets), Healthcare (MRN, diagnosis codes), Communication (email, phone, address), and contextual entities (PERSON, ORGANIZATION, LOCATION) via NER. Includes US state driver's license formats with more international IDs coming soon.
- Does NeuronEdge scan my documents or images?
- No. NeuronEdge operates at the API layer—the last mile before your data reaches LLM providers. By the time requests reach us, your RAG pipelines have already retrieved and converted documents to text, and any images have been processed. We protect the text payload in the API request, not your source files.
- Does NeuronEdge protect tool calls in agentic AI workflows?
- Yes, on all tiers. NeuronEdge parses the OpenAI tool_calls format in streaming responses, redacting PII from function arguments, tool results, and parallel tool calls in real-time. This works with any framework that sends tool-use messages—LangGraph, Claude Agent SDK, OpenAI Agents SDK, or custom implementations.
- Can NeuronEdge detect PII in LLM responses?
- Yes. Response-side detection uses a sliding buffer to scan assistant-generated content for PII—including hallucinated names, addresses, and identifiers that never appeared in the original prompt. Available on Professional and Enterprise tiers.
- What compliance frameworks does NeuronEdge support?
- Five frameworks with one-click policy templates: HIPAA Safe Harbor (18 identifiers), PCI-DSS (cardholder data), SOC 2 (trust service criteria), FINRA (financial PII), and FedRAMP (federal sensitive data). Generate audit-ready compliance reports in PDF or JSON with entity breakdowns, performance metrics, and zero-persistence attestation.
- How does NeuronEdge handle multi-team organizations?
- Enterprise customers get workspaces with policy inheritance. Create separate workspaces for teams, departments, or products—each with their own detection policies, API keys, and analytics. Org-level policies cascade down unless explicitly overridden, so compliance standards are enforced globally while teams retain flexibility.
Ready to secure your AI API calls?
Join our private beta and be among the first to experience enterprise AI security.