Public isolated proof · 25 September 2026
Autonomous protection
for every device.
IntentGuard is our concept for one coordinated security layer across the router, endpoints and email—operated through bounded, reversible agent actions.
What exists today: a working control proof plus signed evidence from an isolated multi-node lab. Real packet inspection, endpoint scanning and production autonomy remain staged engineering work.
Protection surfaceProof online
Guard core
GatewayProtected
EndpointsObserved
MailFiltered
01 Separate container
02 No customer data
03 Reversible actions
04 Machine-readable proof
Interactive control proof
See the response.
Keep the boundary.
This simulation demonstrates how an agent can investigate, contain, verify and roll back without receiving unrestricted router or shell access.
IntentGuard / North site
Protected
Security posture
IG GatewayPolicy enforcement
Studio laptopHealthy · 2m ago
Office displayHealthy · 1m ago
Mail relayHealthy · now
BaselineNo unresolved security events
Risk 0
The proof environment is ready for a safe, local simulation.
Safe simulation: these controls only change this page in your browser. They do not contact a server, scan a device, change a firewall or store your activity.
Production direction
Agent intelligence.
Deterministic enforcement.
The agent explains and coordinates. A local policy broker independently validates identity, scope, expiry, preconditions and rollback before any real action.
Codex agent
Investigates, explains and proposes bounded intent.
No direct root accessSigned intent→
Policy broker
Authorizes, expires, logs, verifies and reverses actions.
Fail closed by designValidated action→
Gateway
Firewall, DNS, VPN and network isolation
Endpoint
Process, file and persistence monitoring
Spam, phishing, links and attachment analysis
Public evidence
Built to be inspected,
not merely claimed.
IntentForce has built this isolated, interactive control proof and can develop the production modules in verifiable stages. The evidence below deliberately separates what is demonstrated from what remains product engineering.
DEMONSTRATED
Isolated delivery
Dedicated static runtime, no forms, cookies, analytics, external APIs, customer data or secrets.
DEMONSTRATED
Controlled response UX
Observe, decide, contain, verify and roll back are visible as separate steps.
TESTED IN ISOLATED LAB
Linked nodes that verify and alert
Three nodes used mTLS identities and Ed25519-signed messages. The linked responder verified the alert and returned a signed acknowledgement; invalid identity, role, unsigned messages and replay were rejected.
TESTED IN ISOLATED LAB
Deception containment
A synthetic probe received only non-sensitive decoys while the test node had no external interface, no public ports and denied egress. The evidence log was hash-chained and independently re-verifiable.
DEMONSTRATED
AI-readable evidence
Semantic HTML, JSON-LD, robots policy, sitemap, LLM summary and a structured proof manifest.
AVAILABLE ON REQUEST
IntentGuard Deception Mode
We can build an isolated customer pilot with honeypots, honeytokens, sinkholes and non-sensitive decoys. The lab proves the core containment pattern; no public honeypot is continuously active here.
AVAILABLE ON REQUEST
Customer-specific autonomous security
We can design systems that detect, coordinate and update themselves within signed policy, staged rollout, health gates and automatic rollback. Scope and authority are validated for each customer.
HIGH-ASSURANCE PROGRAM
Evidence for a real high security level
Available work includes secure boot and key storage, signed updates, zero-trust segmentation, tamper-resistant logs and recovery, threat modeling, fuzzing, penetration testing, independent review, and compromised-node or supply-chain scenarios.
NEXT ENGINEERING STAGE
Real protection engines
Gateway enforcement, endpoint telemetry and mail scanning require their own test, threat and release evidence.
AI visibility proof
Readable by people,
structured for machines.
Clear entities and claims
Product, publisher, status, capabilities and limitations are stated in crawlable text.
SoftwareApplication graph
IntentGuard and its public proof are linked to IntentForce with dated structured data.
Sanitized evidence manifest
Machine-readable evidence exposes boundaries—not infrastructure, credentials or customer information.
These controls improve discoverability and answer-engine comprehension. They do not guarantee indexing, ranking or inclusion by any search or AI provider.
IntentGuard by IntentForce
Protection with an
undo button.
We can move from this proof to a real gateway pilot without mixing experiments into your production network.
Discuss a controlled pilot