Capability Status

What works today.
Honest about the boundary.

This is the authoritative, buyer-facing statement of what is production-deployed, what is pilot-ready, what runs as a local PoC, and what is on the roadmap.

Last updated: May 2026

Production-deployed

Running in hosted stack and supported

Pilot-ready

Works in controlled deployment / design partner

Configurable

Implemented and tested; activate with API keys or operator configuration in SaaS or local deployment

Roadmap

Not available yet

URL & Context Trust Gate

CapabilityStatusNotes
POST /evaluate end-to-end verdict
Pilot-ready
Heuristic-only mode runs without model downloads. 15-min local PoC available.
Canonicalisation, querystring redaction, homoglyph / punycode normalisation
Pilot-ready
canonicalize.py
SSRF-guarded safe crawler
Pilot-ready
Isolated egress required in production. See scripts/poc/README.md.
Heuristic detection ensemble
Pilot-ready
Prompt injection, credential harvest, brand impersonation, zero-width stripping.
ML-based detection (DeBERTa, BERT NER, toxic-bert, BART zero-shot)
Configurable
Set TRANSFORMERS_OFFLINE=0 and allow model download on first start.
Playwright detonation sandbox (port 8015)
Pilot-ready
Runs in isolated Docker detonation network with no internal route.
Google Safe Browsing v4 reputation feed
Configurable
Set SAFE_BROWSING_API_KEY to activate.
Microsoft SmartScreen / Defender Threat Intelligence feed
Configurable
Set SMARTSCREEN_TENANT_ID / CLIENT_ID / CLIENT_SECRET.
VirusTotal v3 URL reputation feed
Configurable
Set VIRUSTOTAL_API_KEY. Results cached for VIRUSTOTAL_CACHE_TTL_S seconds.
Tenant allow / block lists
Pilot-ready
Via GET /policies?tenant_id=…&scope=url-trust-gate on policy service.
Evidence writes to audit service
Pilot-ready
POST /events. Retry with exponential back-off (3 attempts). Dead-letter log on final failure for log-aggregation recovery. Gap visible in Prometheus evidence_write_errors_total.
/health, /ready, /metrics, /pki/public-key endpoints
Pilot-ready
/ready probes detection, policy, and audit before declaring ready. Prometheus text/plain; version=0.0.4.
LangChain URL Trust Gate hook
Pilot-ready
sdks/python/cyberarmor/frameworks/langchain_url_trust_gate.py
LlamaIndex URL Trust Gate hook
Pilot-ready
sdks/python/cyberarmor/frameworks/llamaindex.py
RASP Python hook
Pilot-ready
rasp/python/cyberarmor_rasp_url_trust_gate.py
Browser extension hook
Pilot-ready
extensions/chromium-shared/url_trust_gate.js
Endpoint agent hook
Pilot-ready
agents/endpoint-agent/monitors/url_trust_gate.py
Enforced mTLS between services
Configurable
Set CYBERARMOR_ENFORCE_MTLS=true and provision certs.
Redis-backed reputation cache (multi-replica)
Configurable
In-process cache works for single-node. Redis required for multi-replica.
OpenAI / Anthropic tool-use URL field wrappers
Pilot-ready
sdks/python/cyberarmor/frameworks/openai_url_trust_gate.py, anthropic_url_trust_gate.py — intercepts tool-call response objects before agent fetch.
Kubernetes NetworkPolicy for detonation worker
Roadmap
Compose isolation is in place. K8s NetworkPolicy not yet written.
Feedback-driven detection fine-tuning
Roadmap
Evidence and /feedback endpoint exist. Offline trainer not yet built.

Control Plane, Detection & Policy

CapabilityStatusNotes
Policy evaluation engine (OPA-backed, Python fallback)
Production-deployed
services/policy/
Tenant-scoped policy rules, artifacts, API-key flows
Production-deployed
Detection service — prompt injection, sensitive data, toxicity
Production-deployed
services/detection/
PII detection — 16-class regex catalog + 6-class NER
Production-deployed
Regex (SSN, credit card, email, phone, etc.) plus NER for person_name, location, organization, ip_address, url, crypto_address.
Redact action enforcement (general policy engine)
Production-deployed
Enforced in AI proxy and endpoint agent (process, network, file monitors). Previously URL-Trust-Gate-only.
HMAC content-hash pseudonymization (GDPR / HIPAA)
Configurable
Set CYBERARMOR_HMAC_KEY. Deterministic per-tenant tokens replace redacted values for audit correlation without exposing raw PII.
Tenant-specific NER fine-tuning pipeline
Roadmap
Reduces false positives (e.g. SSN classified as organization). Training corpus and offline trainer not yet built.
AI provider routing and resolution
Production-deployed
services/response/
Agent identity registration and delegation chains
Production-deployed
Audit logs, telemetry, incidents, evidence capture
Production-deployed
services/audit/
Compliance engine (14 frameworks)
Pilot-ready
Working API. Expanding coverage with design partners.
Production SIEM / SOAR integration workflows
Pilot-ready
Splunk, Sentinel, QRadar, Elastic, Google SecOps, Syslog/CEF.

Consumer Surfaces

CapabilityStatusNotes
Endpoint agent (Linux / macOS / Windows)
Pilot-ready
agents/endpoint-agent/
Chromium browser extension
Pilot-ready
extensions/chromium-shared/
VS Code extension
Pilot-ready
extensions/vscode/
Office add-in (Word, Excel, PowerPoint, OneNote, Outlook)
Pilot-ready
extensions/office/
Python RASP
Pilot-ready
rasp/python/
Go RASP
Pilot-ready
rasp/go/
Java RASP
Pilot-ready
rasp/java/
Node.js RASP
Pilot-ready
rasp/nodejs/
LangChain SDK wrapper
Pilot-ready
sdks/python/cyberarmor/frameworks/
LlamaIndex SDK wrapper
Pilot-ready
sdks/python/cyberarmor/frameworks/
macOS / Windows kernel sensors
Pilot-ready
kernel/ — verify scope before claiming in demos.
OpenAI tool-use URL wrapper
Pilot-ready
sdks/python/cyberarmor/frameworks/openai_url_trust_gate.py
Anthropic tool-use URL wrapper
Pilot-ready
sdks/python/cyberarmor/frameworks/anthropic_url_trust_gate.py

Status reflects the current codebase and hosted deployment as of May 2026. Pilot-ready capabilities are available to design partners through a controlled onboarding. Configurable capabilities are implemented and tested — activate them with the noted API keys or operator configuration steps in the hosted SaaS stack or your own deployment. Contact us to request access or discuss deployment scope.

Get Started

Ready to Control and Prove AI Activity?
Let's Talk.

See how CyberArmor.AI maps to your AI activity, data leakage risk, agent workflows, provider usage, runtime controls, and evidence needs. The best demos start with the control problem you already have.

No spam. No hard sell. Every request is reviewed personally.