openwebui-mcp-router is where users enter the mesh.
WHAT'S HAPPENING
Its one pod is running and ready. No destination it calls is failing. 0 of 1 MCP servers reachable. 1 had no calls in the last 5 minutes. Every data plane is connected to the control plane.
Pods ready1 of 1
Data planesall healthy
Control plane1 of 1 connected
SPIRE certificateRotating automatically
Ask
Ask Greymatter⤢ ✕
Is openwebui-mcp-router-version-main healthy, and what depends on it?grok-4.20-0309-non-reasoning · tools on
Yes, openwebui-mcp-router (the service matching the query openwebui-mcp-router-version-main) is healthy.
Its health status is stable (sidecar stable, 1 of 1 pods ready), with low business impact. It uses FIPS-compliant mTLS (TLSv1.3/TLSv1.2 on ingresses), SPIRE for identity, and has a CRD-based origin in the openwebui namespace.
What depends on it:
Configured dependents (1 direct): The service openwebui-mcp-everything (healthy, in namespace openwebui).
Transitive: passthrough (1).
Observed dependents (live callers): 3 total (2 at hop 1, 1 at hop 2). No non-running dependents.
A living catalog of every app, API and AI agent in your project.Every connection, verified. mTLS, TLS and post-quantum, with no code changes.See what breaks for your customers, before it happens.Troubleshoot and report using plain language for all of your business apps.Greymatter.io. Built for when the stakes are highest.
Cross-cutting · infrastructure automation & orchestrationPlatform Orchestration→Provision·Reconcile·Failover·Tear Down
Recognition
Proven in the field. Recognized by the industry.
CyberSecurity Breakthrough · 2026
Named Network Control Center Solution of the Year for the second year in a row.
The 10th annual CyberSecurity Breakthrough Awards spotlight how Greymatter.io gives teams command of their apps, APIs, and AI agents with a single platform.
GET/api/kernels/… Would Blockdisclosurepromptinjection
GET/api/kernels/… Would Blockpromptinjectionoutputhandling
KEY EXCHANGEX25519MLKEM768 · POST-QUANTUM
Data exposed in transit.Encrypted. Inspected. Fingerprinted.
An attacker inside your network can read your traffic, your AI prompts, and your logins, or reuse them.Data stays private, even from future quantum computers. Greymatter catches attacks and flags leaks before they do damage.
03See
data exportLIVE
FORENSIC AUDIT · IDENTITIES
LAST 24H
12,486
PEOPLE SIGNED IN
46 namespaces
38,912
WAF EVENTS
36,507 blocked · 2,405 would block · 214 clients
84,317
GUARDRAILS EVENTS
21,904
DLP EVENTS
PEOPLE BY NAMESPACE
payments
3,412
people · 18,240 sign-ins
claims-processing
2,987
people · 15,903 sign-ins
customer-portal
2,654
people · 14,118 sign-ins
fraud-analytics
1,893
people · 9,822 sign-ins
FORENSIC AUDITacting for j.o●●●●●●
No idea who did what.The whole story, with names attached.
When something goes wrong, your team spends weeks piecing together what happened.Trace any action to a person, a time, and the rule that allowed it. Answer the hard question the day someone asks it.
1 / 3 · Swipe for more
The Common Application Picture
1 platform
Fewer tools to buy, run, and patch.
The capabilities work together from the start, in one platform. No integration project. No surprise add-ons.
Common Application Picturecapability 01 of 07 · built in
This request hit 2 different concern areas at once. It asked to look up or disclose a date of birth (Patient: M•••• L••••, DOB). That is solicitation (fishing for sensitive data), not proof that those values were already present in the message. Solicitation checks are often audit-only so you can see the attempt even when traffic is allowed. It included a medical record or health-plan style identifier (MRN ••••2913). That is health-adjacent identity data. This class of finding is often audit-only so teams can see sensitive content moving through this traffic. It included clinical coding (for example ICD-10, CPT, LOINC, or NDC) (ICD-10 E11.9). Health identifiers are audit-only by design here: the goal is visibility, not always an automatic block. What it measured: pii_score=51.
AI agents, tools, and models, checked as they work.
Guardrails check prompts, answers, agent tool calls, and even text inside images, right beside the app. Guardrails block attacks when you choose. DLP flags sensitive data in both directions.
Why it matters
Say yes to AI projects sooner. Greymatter checks agents, tool calls, and model answers where they run, and records who sent them.
Greymatter checks 12 controls on your apps, APIs, and AI agents, then scores them against more than 20 frameworks. These include NIST, ISO, FedRAMP, HIPAA, PCI DSS, SOC 2, GDPR, DORA, and the EU AI Act.
Why it matters
Be ready for an audit any day of the year. The evidence comes from the live system, not spreadsheets and screenshots.
This request hit 2 different concern areas at once. It asked to look up or disclose a date of birth (Patient: M•••• L••••, DOB). That is solicitation (fishing for sensitive data), not proof that those values were already present in the message. Solicitation checks are often audit-only so you can see the attempt even when traffic is allowed. It included a medical record or health-plan style identifier (MRN ••••2913). That is health-adjacent identity data. This class of finding is often audit-only so teams can see sensitive content moving through this traffic. It included clinical coding (for example ICD-10, CPT, LOINC, or NDC) (ICD-10 E11.9). Health identifiers are audit-only by design here: the goal is visibility, not always an automatic block. What it measured: pii_score=51.
AI agents, tools, and models, checked as they work.
Guardrails check prompts, answers, agent tool calls, and even text inside images, right beside the app. Guardrails block attacks when you choose. DLP flags sensitive data in both directions.
Say yes to AI projects sooner. Greymatter checks agents, tool calls, and model answers where they run, and records who sent them.
90% of Zero Trust controls in place across your apps, AI, and workloads
9 of 10 Greymatter controls are in place everywhere they apply. A control counts only when it is running everywhere it should be.
as of Nov 4, 2026 at 01:55 PM EST read 01:55:48 PM EST · every 3s
Greymatter's controlsGreymatter checks them on the workloads it manages.
Every service has its own identity
IN PLACE
Its own certificate, issued by Greymatter.
49 of 49 services
FIPS 140-3 and post-quantum cryptography
IN PLACE
Validated crypto, certificate 5191.
FIPS 140-3 IN PLACE
63 of 63 services
Post-quantum key exchange IN PLACE
36 of 36 services
Traffic between services is encrypted
▲NOT IN PLACE
Encrypted, identified connections only.
34 of 35 services1 service without it
Certificates renew on their own
IN PLACE
No manual steps.
49 of 49 services
Each service limits who may call it
IN PLACE
Only named sources get through.
Running on all 14 services that declare it
Web application firewall
IN PLACE
Web attacks inspected at the edge.
Running on all 13 edges that declare it
Responsible AI guardrails
NOT TURNED ON ANYWHERE
Prompts, outputs, tools, and agents checked. Every exchange recorded.
No services declare it
Sensitive data is detected
NOT TURNED ON ANYWHERE
Personal, health, and classified data flagged.
No services declare it
Forensic audit trail
IN PLACE
The hops, the identities, and the full chain of a request. Searchable, kept as long as you set.
49 of 49 services
Security policy reaches every service
IN PLACE
Delivered everywhere. Drift caught.
49 of 49 services
Every configuration change is logged
IN PLACE
Who changed what, in a tamper-evident log.
On for the mesh
Console access is controlled
IN PLACE
Role-based, from your identity groups.
On for the mesh
FRAMEWORKS
▲DoD Zero Trust5/7
▲CISA ZTMM5/7
▲Federal mandates3/5
DoD Zero Trust Strategy
7 pillars · target level due 30 Sep 2027
UserIn place01Every service has its own identity02Certificates renew on their own12Console access is controlled
DeviceIn place01Every service has its own identity04FIPS 140-3 and post-quantum cryptography
Network & environment▲Not in place03Traffic between services is encrypted05Each service limits who may call it06Web application firewall
Application & workloadIn place05Each service limits who may call it06Web application firewall10Security policy reaches every service
Data▲Not in place08Sensitive data is detected04FIPS 140-3 and post-quantum cryptography
Visibility & analyticsIn place09Forensic audit trail11Every configuration change is logged
Your score against the major frameworks, live.
Greymatter checks 12 controls on your apps, APIs, and AI agents, then scores them against more than 20 frameworks. These include NIST, ISO, FedRAMP, HIPAA, PCI DSS, SOC 2, GDPR, DORA, and the EU AI Act.
Be ready for an audit any day of the year. The evidence comes from the live system, not spreadsheets and screenshots.
Post-quantum encryption from the first request. Data stolen today stays safe from future quantum computers.
One record
Who did what, and which rule allowed it.
j.r●●●●● → recordsLOGGED
billing → recordsDENIED
Faster audits
62%
Faster audits for one customer.
Fewer tools
WAFGWAIGMESHAUDIDOPS
Several capabilities, one install.
AI coverage
10/10
OWASP risks for AI apps, covered.
Validated
FIPS 140-3
FIPS 140-3 validated (CMVP #5191). The encryption standard federal buyers require. NIST post-quantum encryption (ML-KEM, FIPS 203) is also on by default.
Greymatter is a Zero Trust runtime security platform for apps, APIs, and AI agents. It contains attackers inside the first system they reach. Your data stays private, even from future quantum computers. In addition, actions are on record: who did what, and when. Greymatter runs wherever your mission does: in the cloud, on premises, or on fully offline networks. It supports Kubernetes, OpenShift, RKE2, k3s, and k0s.
Greymatter delivers Zero Trust runtime security, which protects your apps, APIs, and AI agents while they run, not only at login. It confirms who is calling on each request. Connections stay private. Meanwhile, the WAF and AI Guardrails catch attacks on the way in. The record names the person behind each step.
Greymatter gives an AI agent its own identity and its own rules, so it can only do what you allow. It screens what goes into and comes out of your AI. The record names the person the agent acted for.
Defense, national security, and civilian government programs use Greymatter. The platform is also built for private sector teams that run systems that cannot fail. Greymatter is FIPS 140-3 validated (CMVP #5191), and post-quantum encryption is on by default. Defense buyers can also procure it through the Tradewinds Solutions Marketplace.
Common Application Picture
capability 01 of 07 · built in