IntentGate Product Brief (v1.9)
the customer-facing introduction in PDF form.
Knowledge center
Agent Runtime Authorization Knowledge Center
The conceptual model, seven runtime controls, standards mappings, product references, operator documentation, and executive material for teams evaluating or deploying the AI runtime security platform and MCP firewall for AI agents.
Jump to a chapter
Resource library
A guided route through IntentGate.
This is not a flat sitemap. Each chapter is sequenced around how security, architecture, compliance, and operations teams actually evaluate the category, from definitions, to controls, to standards evidence, to implementation.
Category and concept
Category and concept
Definitions and the broad picture for the agent runtime authorization control category.
What is Agent Runtime Authorization?
The canonical definition of the category: what it is, the threat model, the seven required controls, how it fits with your existing security stack. Start here.
Glossary
60+ definitions covering the agent runtime authorization vocabulary: capability tokens, intent enforcement, destructive verbs, memory provenance, MCP, Rego, and related concepts.
The Seven Controls
The Seven Controls
Deep-dive pages on each of the seven controls a complete agent runtime authorization solution must implement.
Capability Tokens
HMAC-bound credentials that scope an agent to specific tools, tenants, and time windows. The foundational control.
Agent Microsegmentation
Zones on the token, default-deny agent-to-agent (east-west) calls, and per-zone tool scope, with a live flow map and estate matrix. Contains lateral movement across the estate.
Intent Enforcement
The gateway control that matches the agent's resolved tool call against the user's declared structured intent. The primary defence against prompt injection.
Policy Engine
Rego-based policy for destructive-verb deny-lists, bulk-row ceilings, value thresholds, and approved-destination lists.
Budget Tracking
Per-tenant token and cost ceilings, fail-closed. The defence against unbounded consumption.
Memory Provenance
Signed chain-of-custody on agent memory writes so high-stakes reads can require verified provenance. Defeats memory poisoning.
Bidirectional PII Filtering
Content inspection on prompts and responses. Nine checksum-validated PII classes plus eight credential and secret classes, with counts-only audit so matched values are never persisted.
Complementary capabilities
Complementary capabilities
Capabilities that work alongside the seven-check pipeline to harden secrets and watch behaviour over time.
Credential Brokering
The gateway holds each tool's credential and injects it per call, so agents never possess a production secret. Encrypted at rest, rotated centrally, enforced as defence in depth against bypass.
Behavioural Anomaly Detection
Baselines each agent over the audit trail and flags denial bursts, new tools, volume spikes, and off-hours activity. Alerts through a webhook and can revoke the offending token automatically.
Standards and threat models
Standards and threat models
How agent runtime authorization maps to the frameworks security teams, auditors, and procurement reference during evaluation.
Standards Alignment
Per-standard mapping for OWASP Top 10 for LLM, OWASP Top 10 for Agentic AI, NIST AI RMF, MITRE ATLAS, EU AI Act, ISO/IEC 42001, SOC 2, ISO 27001, GDPR.
OWASP Coverage
The CISO-readable OWASP coverage matrix. 14 risks directly mitigated, 1 partial, 5 explicitly out of scope by deployment model.
Product
Product
What IntentGate is and how it works in practice.
Why IntentGate
The four-control bypass essay. The sharpest single positioning page on the site.
Product
What ships, what it does, where it fits.
See it work
Live evaluation lab: a representative agent deployment with all controls enabled and audit chain visible end-to-end.
How it works
Five-check authorization pipeline walkthrough with the install demo cast.
Security & Compliance
Tamper-evident audit, RBAC, JIT elevation, evidence for SOC 2, ISO 27001, GDPR, AI Act.
Open Source
Apache 2.0 licensed gateway, contribution model, the relationship between the open-source core and IntentGate Pro.
Documentation
Documentation
Operator-facing reference for deploying and operating IntentGate.
Executive briefings
Deeper material for evaluation and procurement.
The following PDFs are available on request through contact . They are the deeper deliverables that complement the public pages: the OWASP-anchored vendor security pack, the CISO-facing threat stories doc, the operations overview, and the architectural comparison battle cards.
IntentGate Threat Stories
the CISO walk-through organised around the CIA triad, with the OWASP-anchored coverage map.
IntentGate Vendor Security Pack (v1.1)
deep technical pack with full OWASP coverage matrix.
IntentGate Operations Overview
CISO-facing operational commitment doc.
IntentGate vs. Zscaler & Microsoft Defender
architectural comparison for prospects evaluating IntentGate alongside their perimeter stack.
IntentGate vs. Azure API Management
architectural comparison for prospects evaluating IntentGate alongside their existing API gateway.
Need the right path for your team?
Tell us whether you are evaluating architecture, control coverage, compliance evidence, or deployment. We will point you to the shortest useful route.