OpenAI
Anthropic
Google Vertex AI
Google Gemini
Grok (xAI)
DeepSeek
Perplexity
Mistral
Cohere
Amazon Bedrock
Hugging Face
Databricks
Azure OpenAI
Qwen
Groq
Azure Monitor
GitHub Copilot
AWS CloudTrail
Slack
Salesforce
Microsoft 365 & Copilot
Snowflake Cortex
Windsurf
Notion
Replit
Microsoft 365 Copilot
Microsoft Entra ID
Microsoft Teams
Okta
Google
Claude Code
Cursor
Kiro
Ollama
LM Studio
Microsoft Sentinel
Palo Alto Networks
Google SecOps
BMCInsurance
Claims triage agent
Amazon BedrockClaude
ClaudeA triage agent reads claims and calls claims-system tools through MCP. Writes wait for a person, and a tool that changes after approval breaks the chain.
ASIC RG 271 · Privacy Act · OWASP LLM06AI trust · whole estateSecurity · Compliance · Cost, in one score
Govern every AI across your entire estate.
Most platforms watch one slice of your AI. Kapālins covers the whole estate — every model, copilot and agent, on Bedrock, Vertex, Azure OpenAI or a developer’s laptop — proves each one is acting exactly as you approved, and scores it for security, compliance and cost as a single AI Trust Index.
See · Protect · Prove 15 AI providers · 69 device tools 29 frameworks · 9 jurisdictions
Live · AI Command Centredemo tenant · 1,284 AI systems
See · shadow AI14 people used 3 AI apps nobody approved this week.Found through Entra grants and the device sensor.
Investigate →0 / 100
AI Trust Index · medium
▲ +0 this week · 0 threats stopped today
Agentic runtimelive
0 agents
behaviour + identity guarded
AI securityOWASP
0 blocked
injection · DLP · jailbreak
Compliance29 frameworks
0%
NIST · APRA · EU AI Act
FinOpson budget
$0k
token spend · DoW guard
Guardrail stream · real time0 / min
0 policies enforcing122+ framework controls mapped0 evidence items
Whole-estate coverage
Every AI your enterprise runs.
AI arrives three ways: the AI you build, the AI already inside tools you license, and the AI nobody approved. Kapālins governs all three, and every logo below is something it connects to today.
AI you buildApps and agents on your models
Bedrock, Vertex, Azure OpenAI and direct APIs, governed inline at the gateway.
AI you already licenseCopilots inside your tools
Microsoft 365 Copilot, GitHub Copilot, Salesforce, Slack and more, read from their own audit logs.
AI nobody approvedShadow AI
Found through identity grants and the device sensor, then brought under policy or blocked.
AI providers15 · inline
Routed through the gateway, with policy enforced on every call. OpenAI and Anthropic dialects.
OpenAIAnthropicGoogle Vertex AIGoogle GeminiGrok (xAI)DeepSeekPerplexityMistralCohereAmazon BedrockHugging FaceDatabricksAzure OpenAIQwenGroq
Counted from the gateway’s provider listAudit-log sources15 · connector
What each platform’s own logs say about AI use: 13 pulled on a schedule, 2 pushed by webhook.
Azure MonitorOpenAI AdminGitHub CopilotAWS CloudTrail
Slack
Salesforce
Microsoft 365 & CopilotSnowflake CortexWindsurfNotionReplit
Counted from the ingestion runner and its webhooksSlackSalesforceMicrosoft 365 & CopilotSnowflake CortexWindsurfNotionReplitMicrosoftCopilot · Entra · Azure
Microsoft 365 Copilot audit and seat usage, SharePoint oversharing, sensitivity labels, Copilot Studio routing and Azure OpenAI inventory.
Microsoft 365 CopilotCopilot StudioMicrosoft Entra IDAzure OpenAIMicrosoft Teams
Microsoft insights and the Copilot Studio recipeIdentity31 apps · grants
OAuth grants matched against a catalogue of AI apps, which finds shadow AI. SCIM 2.0 keeps people in step.
Microsoft Entra IDOktaGoogle
SCIM from all four · grant discovery from Entra and OktaOn the device69 tools · sensor
A desktop sensor for Windows and macOS and a browser extension for Chrome, Edge and Firefox. It finds local AI and gates prompts, files, MCP tools and egress.
Claude CodeCursorWindsurfGitHub CopilotKiroOllamaLM Studio
Counted from the device catalogueSecurity operations7 · push
Findings and blocks pushed to the tools your SOC already runs, plus signed webhooks.
Microsoft Sentinel
Palo Alto NetworksGoogle SecOpsBMCFive ways in
Wherever AI runs, there’s a way in.
In the traffic, the tests, the logs, the logins and the laptops. A gateway alone sees one of these. Kapālins uses all five, so the AI your teams actually use is governed, not just your API traffic.
Mode 01the traffic
Proxy
The traffic. An inline gateway with full payload visibility: allow, redact, hold or block every call.
Mode 02the tests
Probe
The tests. Red-team probes and a firewall check run against each AI endpoint, so behaviour is tested rather than assumed.
Mode 03the logs
Connector
The logs. Audit logs from each platform’s admin API show what is authorised, configured and drifting.
Mode 04the logins
Identity
The logins. OAuth grants are matched against 31 AI apps, so AI someone approved with one click is found.
Mode 05the laptops
Device
The laptops. A sensor finds 69 AI tools and gates browser prompts, files, local MCP tools and egress.
Shadow AI
Bring the AI nobody approved into the light.
Personal ChatGPT accounts, coding agents on laptops, AI apps approved with one click, a tool server holding a production password. Kapālins sees all of it, says what happened to every interaction, and puts each tool in front of a person to decide.
Demo estate. The layout and every state mirror the platform’s Shadow AI page. How shadow AI is found
Autonomous Integrity
Is every AI still acting exactly as you approved?
Each application is a chain: the agents and credentials that act for it, and the models, tools and partner agents they reach. Six plain questions are asked of every part, and the hexagon shows the answers. Click any application, tile or question to open it.
How Autonomous Integrity works
Demo estate, drawn with the same seal, map and wording as the platform.
The whole platform
41 capabilities. 12 pillars. One score.
Every node is a capability that ships today, counted from the platform code. Select a node, switch the lens to a framework or a persona, or run any of the 34 scenarios and tick them off as you go.
41capabilities12pillars34scenarios1score
One number for the board
The AI Trust Index.
Boards cannot read a forty-tab dashboard. The AI Trust Index is a single 0–100 score — like a credit rating for your AI estate — composed of three weighted pillars, every input traceable to evidence.
AI Trust Index
0
● MEDIUM · 65–84
▲ +11 vs last quarter
AI Security weight 50%86 / 100
Identity, data, model and usage risk, measured across all five ways in and the assessment.
AI Compliance weight 30%70 / 100
Every applicable framework, each control proved by a live signal or the assessment.
AI Value weight 20%40 / 100
FinOps cost attribution and token-budget fences — the CFO’s question, answered beside the CISO’s.
Example tenant. Composite 0–100 per tenant. Bands: 85+ good · 65–84 medium · 45–64 high risk · <45 critical. How the AI Trust Index works.
Use cases
The AI regulated industries actually run.
Six applications you will recognise, the controls Kapālins applies to each, and how each one reads on the integrity view above.
All staff
Microsoft 365 Copilot
Microsoft 365 Copilot
Copilot can only be as safe as what it can read. Oversharing in SharePoint and sensitivity labels are checked, and seat usage shows who actually uses it.
Privacy Act · ISO/IEC 42001 · OWASP LLM02Marketing & sales
ChatGPT Enterprise
ChatGPT Enterprise
The OpenAI admin connector shows who used which custom GPTs, and personal ChatGPT accounts on company devices are found and flagged.
GDPR · Privacy Act · OWASP LLM02Engineering
Claude Code
Claude CodeGitHub
GitHubCoding agents on developer laptops. The device sensor finds them, gates local MCP servers and egress, and holds secrets back from prompts.
NIST SSDF · EU CRA · OWASP LLM02Telco & retail
Customer service agent
Google Vertex AIGemini
A customer agent books through partner agents. Agent cards are checked by signature, and an agent nobody approved shows as unapproved.
GDPR · PDPA · OWASP LLM06Banking
Credit memo assistant
Amazon BedrockClaude
Drafts credit memos from policy search. Every call goes through the prompt firewall, identifiers are redacted, and CloudTrail proves which model answered.
APRA CPS 234 · Privacy Act · OWASP LLM02Compliance
29 frameworks. 122+ controls. Evidence, not a binder.
Each control is tied to the assessment questions and the live signals that prove it. Pick a jurisdiction, then a framework, to see its controls and how each is evidenced. The Middle East frameworks are available too.
evidence chain
Open the full framework explorer
Regulators are shown by flag and name. Their emblems are not licensed for vendor use.
Why us
Capability. Capacity. Coverage.
We win on the three things that matter to a regulated enterprise: how much of your AI we can see, how deeply we can govern it, and how completely we can prove it.
Whole-estate coverage
Five ways in reach every model, copilot and agent, not just the API traffic a gateway can see.
One board number
Security, compliance and cost composited into a single AI Trust Index your board can act on.
Security, five ways in
Inline enforcement, scheduled probing, audit-log inventory, identity grants and the device sensor, plus the assessment.
APAC-native regulation
APRA CPS 234, DPDP, the Privacy Act, MAS, BNM RMiT and Japan’s APPI, each mapped to the platform and proved by a live signal or the assessment. The Middle East frameworks are available too.
Hosted in Australia
The platform and its data are hosted in Australia. How we handle personal information is set out in our privacy policy.
AI FinOps
Cost attribution per tool, team and tenant — beside the security and compliance view.
Partners & MSSP
Run Kapālins as your managed service.
A partner console of its own, each customer isolated in its own tenant, Guard Unit pools with settlement and margin, and evidence packs issued under your name.
Programme
Authorised
Regional MSSP, SI practice- Co-branded with Kapālins
- 2 sandbox customers
Programme
Premier
Established security practice- White-label eligible; client software says “Powered by Kapālins”
- 5 sandbox customers
- Managed policy for every customer
- Customer-owned branding
Programme
Elite
Carrier, global SI, national MSSP- White-label included; client software says “Powered by Kapālins”
- 5 sandbox customers
- Managed policy for every customer
- Customer-owned branding
- Dedicated database
- Unlimited team seats, and data regions as they open
Services
Help getting live, from the people who built it.
Kapālins is a product. These services exist to get it working on your estate quickly.
AI adoption roadmapFrom inventory to policy to scale, on your estate.
Onboarding & implementationGateway, connectors, sensor and SSO, switched on with you.
AI red teamingInjection, agents, MCP and A2A, run with the platform’s own probes.
By the numbers
Procurement-grade from day one.
Product facts, each counted from the platform code and regenerated when the platform changes.
15
AI providers through the gateway
32
Policy Shield rules in 7 families
29
frameworks in the crosswalk
122+
controls tied to evidence
Resources
Go deeper.
Autonomous IntegrityIs every AI acting as you approved?Six questions asked of every part, and a passport anyone can check.Framework libraryEvery framework and the evidence behind itEach control and the signals that prove it.InsightsRegulatory radarWhat was published and what takes effect next, from the crosswalk.AssessmentsFind out where you standFive assessment categories, an executive pack at the end.
Request a briefing
Govern your AI before your regulator asks.
A 30-minute walkthrough of the Trust Index, the five ways in and your regulator’s evidence chain — on your estate.
FAQ
Questions buyers ask.
What is Kapālins?
An Australian-owned enterprise AI security and governance platform. It finds every AI an organisation runs, enforces policy inline, proves each application is acting exactly as it was approved, and turns compliance into evidence. Security, compliance and cost roll up into one AI Trust Index.
How is Kapālins different from an AI gateway?
A gateway only sees the traffic routed through it. Kapālins has five ways in: the gateway, scheduled probes, audit-log connectors, identity grants and a device sensor. That is how it governs Copilots, coding agents and shadow AI that never touch a gateway.
Which AI providers do you cover?
15 providers through the gateway, including OpenAI, Anthropic, Google Gemini and Vertex AI, Amazon Bedrock, Azure OpenAI, Grok, DeepSeek, Mistral, Qwen and Groq. 15 audit-log sources, including Microsoft 365 Copilot, GitHub Copilot and AWS CloudTrail. 31 AI apps found through identity grants, and 69 AI tools found on devices.
What is Autonomous Integrity?
It answers one question for each AI application: is it still acting exactly as a person approved? Each application is drawn as the chain it acts through, every part is asked six questions, and the application is rated by its weakest link, never an average.
Which compliance frameworks are covered?
29 frameworks in 9 jurisdictions plus global standards, with 122+ controls tied to evidence. The Middle East frameworks, including Saudi PDPL, SAMA CSF, NCA ECC and UAE PDPL, are available and mapped to your controls at onboarding.
Are you SOC 2 or ISO 27001 certified?
Not yet. SOC 2 Type II and ISO/IEC 27001 are in progress, and we will say so here when each is certified.
Can an MSSP run Kapālins for its customers?
Yes. The partner programme has three tiers: Authorised, Premier and Elite. Partners get their own console, isolated customer tenants, Guard Unit pools and co-branded or white-label delivery. Client software always says Powered by Kapālins.
Where is our data hosted?
Hosted in Australia. Our privacy policy explains how personal information is handled.