THE PLATFORM
Complete AI Security Suite
Runtime protection, agent governance, protocol security, and data loss prevention. Eight products, one platform, under 100ms.
Request Demo
ENFORCEMENT AND IDENTITY
DelphiGateway
An enforcement point in front of your agents. Every call to a model, a tool or another agent is decided before it reaches its destination, with the principal behind it verified against your own identity provider.
Inline enforcement before execution, not after the fact
Verified principal on every request: person, application or autonomous workload
On-behalf-of delegation chains reconstructed across hops
Per-principal quarantine that takes effect on the next request
Model and tool entitlements enforced per agent and per deployment
Bring your own identity provider over OIDC, validated offline
Legacy and third party agents covered with no code changes
Self-hosted in your environment, air-gapped if required
ARCHITECTURE
One control. Five layers.
Delphi enforces a single rule: an authorised action must trace to an authorised intent. Each layer contributes findings, and every finding feeds one trust score that attenuates what an agent is permitted to do.
L1 Text detection. Inspects the language itself: prompt injection, jailbreaks, obfuscation and sensitive data, with over 300 rules across threat families.
L2 Action semantics. Parses commands and tool calls into structure and classifies them by impact and tier, so what an action operates on matters as much as the verb.
L3 Virtual patching. Push targeted protections for a newly disclosed attack class to running agents without waiting for a release.
L4 Behaviour. Measures every agent and principal against its own history, surfacing rarity, trajectory and drift.
L5 Intent verification. Traces an action back to the request that authorised it, so an action with no authorising intent is visible as such.
AGENT GOVERNANCE
xAIDR
Extended AI Detection and Response for multi-agent systems. Fleet management, A2A content inspection, shadow AI detection, and compliance enforcement at every node.
Per-agent trust scores and behavioral baselines
Agent-to-agent runtime content inspection
Shadow agent detection and quarantine
Cross-session correlation and threat intelligence
Microsoft AGT integration for policy enforcement
Enforcement mode toggle: blocking or monitor
PROTOCOL SECURITY
MCP Shield
Security checkpoint for Model Context Protocol communications. Tool call interception, server authentication, and least-privilege enforcement.
Tool call payload scanning before execution
267 detection rules applied to MCP traffic
Server identity verification
Blocked tools and blocked URLs enforcement
Works with any MCP client or server
One-line integration via SDK
DATA PROTECTION
Enterprise DLP
Bidirectional data loss prevention for all AI traffic. Scans prompts and responses with multi-classifier policy enforcement. Patent pending.
15 PII classifiers with custom policy rules
Bidirectional scanning of prompts and responses
Context leakage prevention
Output scanning with strict, moderate, and permissive modes
Custom block and allow patterns per deployment
Full audit trail with zero conversation storage
PIPELINE SECURITY
RAG Protection
End-to-end scanning for retrieval-augmented generation pipelines. Detects indirect injection, knowledge base poisoning, and context manipulation before they reach your model.
Indirect prompt injection detection in retrieved documents
Knowledge base poisoning detection
Document-level threat scoring
Dedicated RAG scan mode with context-aware rules
Integration with any vector database or document pipeline
Validated across 3,900+ adversarial test payloads
CRYPTOGRAPHIC IDENTITY
Agent Identity
Cryptographic identity infrastructure for AI agents. Every agent gets a verifiable identity, every request gets signed, every action is attributable.
Ed25519 keypair generation per agent
Cryptographic request signing on every API call
Decentralized Identifier (DID) support
Identity persistence across sessions and restarts
Identity-based policy scoping per agent
Tamper-evident audit chain
COMPLIANCE AND GOVERNANCE
Compliance
Automated compliance mapping, assessment, and evidence collection across every major AI governance framework. Runtime evidence, not annual audits.
OWASP LLM Top 10 detection mapping
NIST AI RMF automated assessments
ISO 42001 compliance scoring
EU AI Act readiness reporting
MITRE ATLAS threat mapping
Runtime evidence export for SOC 2 and ISO 27001 audits
TAILORED TO YOUR ENVIRONMENT
Customization
Every deployment is different. Delphi learns your environment, adapts its detection profile, and tunes itself to your domain.
Observation Mode
Two-week observation period captures your real traffic patterns, topic distribution, and false positive patterns before any tuning begins.
Domain-Aware Profiles
Custom security profiles built from your observed traffic. Blocked topics, sensitivity thresholds, and detection rules tuned for your specific use case.
Continuous Tuning
Monthly traffic reviews, threshold adjustments, and profile updates. Your detection improves with every deployment cycle.
Self-Hosted Deployment
Full platform deployed in your VPC or on-premises. No data leaves your environment. Delivered as a Docker container with your custom profile baked in.
FAQ
Common questions.
How does Delphi work with my existing LLM provider?
One URL change points your existing API calls through the Delphi proxy. No provider changes, no code rewrites. Works with OpenAI, Anthropic, Google, Cohere, Mistral, and any OpenAI-compatible endpoint.
How long does deployment take?
Under 10 minutes for the proxy integration. The npm sensor install takes under 5 minutes. Most teams are running in production the same day.
What is xAIDR and do I need it?
xAIDR is Delphi's fleet security layer for multi-agent systems. If you run a single LLM application, AI Sentinel is sufficient. If your agents communicate with other agents, xAIDR adds inline A2A scanning, trust scoring, and shadow agent discovery.
What compliance frameworks does Delphi map to?
OWASP LLM Top 10, OWASP Agentic Top 10, OWASP MCP Top 10, NIST AI RMF, EU AI Act, ISO 42001, AIUC-1, and AWARE. Evidence is generated automatically from runtime detection data. No manual questionnaires.
Is there a free trial?
Yes. 14 days, full access, no credit card required. The free tier includes 15,000 API calls per month with no time limit after the trial ends.
GET STARTED
See Delphi in action
30-minute demo tailored to your AI agent stack.