AI & the Kappa Data portfolio

We don’t sell the AI.
We make it valuable & safe.

Your customers pick Claude, Copilot, Mistral, Gemini or ChatGPT themselves. The value is in what that AI can reach, what it is allowed to see and who is accountable for its actions. That is the security, network and identity stack we sell, integrate and support.

Which AI does your customer use?

Your AI
platform
Claude · Copilot · Mistral · …
governed by Kappa Data
ConnectMCP to your stack ProtectPrompt inspection & DLP Secure APIsWhat agents call GovernPeople & agents
Built on the vendors we distribute, integrate and support + runZero · ExtremeCloud IQ · Sophos · SolarWinds
Our value on top of the model

The model is the engine.
We build the roads, locks and licence plates.

Every AI vendor sells intelligence. None of them can see the other AI tools your people use, connect safely to your whole multi-vendor stack, or give agents an identity your auditors accept. That gap is where Kappa Data comes in.

What you get from the AI platform alone — and what we add

Customer questionClaude alone+ Kappa Data stack
Can the AI answer questions about our network, identities and vulnerabilities?~Only if someone wires up connectors✓MCP to runZero, JumpCloud, Cato, XIQ …
Who may use which connector, and with whose credentials?–API keys per user, in the AI client✓JumpCloud AI Gateway: per group, per tool, OAuth
Do we see all AI tools in use, including shadow AI?–Only its own app✓Barracuda & Cato discover every GenAI app
Can we stop a card number or source code from being pasted into a prompt?~Inside its own tenant, if licensed✓Inline prompt inspection & DLP across every AI tool
Are the APIs our agents call protected against abuse?–Out of scope for the model✓Barracuda API Protection: discovery, bots, DDoS
Does every agent have an owner, an identity and an expiry?–Agents run on shared service accounts✓JumpCloud Agentic IAM
Can we prove it to an auditor (NIS2, EU AI Act, ISO 27001)?~Fragmented logs per tool✓Central tool-call & prompt audit logs, access reviews
01 · Connect — Model Context Protocol

Ask your whole stack
one question.

MCP lets an AI assistant talk directly to vendor platforms. You ask in plain language, the assistant picks and calls the right tools itself. Pick a scenario and watch it run through the JumpCloud AI Gateway.

Claudeconnected toJumpCloud AI Gateway Simulation based on real lab sessions
Ask anything about your environment…
The enterprise challenge

Every user × every MCP server
doesn’t scale. One gateway does.

Drag the sliders to size your customer. Without a gateway every user stores API keys for every server in their AI client, with no central overview.

100
connections to configure
100
API keys living in AI clients
4
places to look for audit logs

    How the gateway works, in the JumpCloud console

    The user authorises their tools

    • Sign in through the JumpCloud User Portal (OAuth)
    • Only MCP servers allowed for the user’s group are shown
    • The user ticks which tools the AI client may use
    • Vendor credentials stay in JumpCloud, never in the AI client
    JumpCloud User Portal where a user authorises MCP toolsEnlarge

    Ask questions through one connector

    • The AI client sees one connector: JumpCloud AI Gateway
    • The gateway routes to the right tool, e.g. runZero query_inventory
    • Every tool call still asks the user for confirmation
    AI client using the JumpCloud AI Gateway connectorEnlarge

    Full visibility for IT & security

    • User identity, AI client, MCP server and tool
    • Status and timestamp — including failed calls
    • Evidence for NIS2 and EU AI Act logging requirements
    JumpCloud AI Gateway activity logEnlarge

    One dashboard for people and agents

    • MCP servers governed by the gateway
    • Active identities: users and agents
    • Tool-call volume, success rate, top servers and tools
    JumpCloud AI Gateway dashboardEnlarge
    No MCP server? Build your own

    From API docs to a governed tool in days

    1. API docs — the vendor’s OpenAPI spec or REST documentation
    2. Generate — an AI coding agent writes the MCP server (Claude Code, GitHub Copilot, Google Antigravity, Mistral Vibe)
    3. Test — connect it to your AI client with real questions
    4. Share — publish it centrally behind the JumpCloud AI Gateway

    Example: our own MCP server for ExtremeCloud IQ answers “which SSIDs are live in De Pinte?”

    Self-built ExtremeCloud IQ MCP server in an AI clientEnlarge

    ⚠︎ Community MCP servers (e.g. for Sophos Central or Barracuda CloudGen Firewall) are often unvalidated. We review the code, the credentials they need and their write actions before use.

    02 · Protect — AI Security

    See every AI tool.
    Stop sensitive data before it leaves.

    Employees paste customer data, credentials and source code into whatever AI tool is closest. Barracuda and Cato inspect prompts and uploads inline, across every GenAI app — not just the one you licensed.

    63%of organisations lack an AI governance policyIBM Cost of a Data Breach 2025
    62%are already experimenting with AI agentsMcKinsey, The State of AI 2025
    $4.44Mglobal average cost of a data breachIBM Cost of a Data Breach 2025

    Prompt inspection lab

    Runs locally in your browser — illustrates the policy “Protect-Sensitive-Data-in-AI”

    1 · Which AI tool is the user on?

    …Waiting for a prompt
    Pick an AI tool and send a prompt. The inspection engine checks it inline, before it reaches the AI provider.
    TimeUserAI toolActionDispositionPrompt
    The prompt audit log fills as you send prompts.
    From our own lab — Barracuda SecureEdge

    This isn’t a slide. It’s running.

    One user, one afternoon in our KD Testing tenant: 47 AI requests discovered, Microsoft Edge Copilot and ChatGPT blocked, Gemini allowed with prompt inspection — and every prompt containing a bank card number denied and audited.

    Barracuda NetworksBest for lean IT teams

    Barracuda AI Security

    See, guide, accelerate: make AI adoption visible, governed and safe — delivered inside BarracudaONE, without enterprise minimums.

    • Shadow AI discovery with a risk classification per AI service
    • Approve, block or guide users to sanctioned tools
    • AI Data Security: real-time inspection of prompts & uploads
    • Prebuilt detectors: PII, credentials, source code, payment cards, health data
    • Block or warn before data reaches the GenAI tool
    • Exportable, audit-ready evidence for insurers and auditors
    • Bailey AI assistant to build policies in plain language
    Cato NetworksBest for SASE customers

    Cato AI Security

    One cloud-native platform for how employees use AI, the AI apps you build and the agents you run — part of the Cato SASE Cloud.

    • Shadow AI dashboard and a catalog of 950+ GenAI applications
    • GenAI DLP with ML-based classifiers, including source code
    • Inspection of prompts and responses with real-time guardrails
    • Protection of home-grown AI apps against prompt injection & jailbreaks
    • Discovery of local agents and connected MCP servers
    • Runtime protection for prompts, outputs and tool calls
    • Same policy engine for every site, user and device
    03 · Secure — API Protection

    Agents don’t click.
    They call APIs.

    Every MCP tool, every agent action and every AI integration ends up as an API call. More agents means more traffic, more endpoints — and more forgotten “zombie” APIs. Barracuda API Protection discovers and defends them all.

    Live API traffic (simulated)
    0
    requests
    6
    APIs discovered
    0
    attacks blocked
    Barracuda NetworksPart of Barracuda Application Protection

    What stops what

    04 · Govern — Agentic IAM

    Agents are identities too.

    With JumpCloud, an AI agent gets its own identity — with a human owner, its own authentication, group-based access and a full lifecycle. See every agent. Govern every identity.

    JumpCloud Click through the lifecycle of an agent

    People vs. agents, one rulebook

    Users, devices, certificates and agents follow one set of policies, groups and audit logs. Disable an agent without touching its human owner.

    No more zombie agents

    Time-boxed access requests and recurring access reviews make sure agents lose access the moment they’re no longer needed.

    Act, not only read

    JumpCloud exposes IAM, PAM, MDM and assets via MCP: onboard, offboard, change access or lock a device from a conversation — with confirmation per call.

    Sales playbook

    What is your customer
    worried about?

    Tick what you hear in the meeting. We’ll map it to the right play, the products and a sentence you can use on the spot — then copy the summary into your follow-up mail.

    Recommended plays 0

    Tick one or more customer concerns.

    Why it matters

    More value from the stack.
    Less cost around it.

    Fewer tools to buy

    IdP, MDM, PAM, PKI, RADIUS and AI governance in one platform instead of separate licences.

    Fewer integrations

    One governed MCP connection instead of point-to-point integrations, API keys and scripts per tool.

    Less time per task

    Troubleshooting, onboarding and reporting in one question instead of hopping between consoles.

    Lower compliance cost

    Logging, access reviews and attestation reports built in — audit evidence in minutes, not weeks.

    How Kappa Data helps

    From first workshop
    to governed production.

    LEARN

    Workshops

    Hands-on sessions on MCP, AI security and agent governance with your stack.

    CONNECT

    Integrations

    Vendor MCP connections and the JumpCloud AI Gateway, set up and tested.

    BUILD

    Custom servers

    MCP servers on top of vendor APIs that don’t have one yet.

    SECURE

    Vetting & policy

    Community-server reviews, read-only scopes, prompt-inspection and DLP policies.

    In short

    One connection. Access per group. Every prompt and tool call visible.

    Want to see it live? We demo the full story end to end — MCP, prompt inspection, API protection and agent identities — on your use cases.