Kyvvu

For security teams

Say yes to agents in production.

The business wants agents; you're asked to sign off on software that decides for itself what to do next. Kyvvu gives you a control that acts before every action, and a record your SOC can work with.

inbox-agent · task 3c60kyvvu engine · in-process
  1. step.execpython export_contacts.pyblock
Blocked: outside content entered this task, so execution needs a fresh approvalTry it: click the highlighted step to take it off the path

The problem

A new kind of insider, with no playbook.

  1. 01

    Real credentials, machine speed

    Agents hold real credentials and act at machine speed. By the time a log line reaches the SOC, the action has happened.

    after-the-fact detection

  2. 02

    Sign-off without a control

    You're asked to approve agents whose behaviour depends on a prompt and on whatever content they read. A design review can't cover that.

    go-live approvals

  3. 03

    No incident process for agents

    What does an agent incident look like, who owns it, and what do you look at first? Most teams haven't rehearsed it.

    incident response

In practice

What your SOC would see, and what never ran.

An email turns the agent against you

An inbox agent triages incoming mail. One message hides an instruction to forward the mailbox to an outside address.

inbox-agent · task 3c60kyvvu engine · in-process

Illustrative scenarios. Each one is decided by a single policy from the open policy library, as the engine evaluates it.

How Kyvvu helps

A control that acts before the step runs.

  1. 01

    Judges the whole path

    Every action is checked against everything the agent did earlier in the task, and blocked before it runs if the sequence is dangerous.

    allow/warn/block

  2. 02

    Incidents your team can follow up

    Every warn or block opens an incident with the policy, its severity, the step and the agent's risk level.

    incidents

  3. 03

    Into the tools you already run

    Traces ship to the SIEM you already run on Enterprise & Public, so agent activity sits next to the rest of your security data.

    SIEM export

  4. 04

    Standard packs to start from

    OWASP agentic, the EU AI Act, GDPR, DORA and MiFID II, healthcare and data exfiltration packs from the open library, next to your own rules.

    open policy library

  5. 05

    Runs inside your environment

    The engine sits in the agent's own process: so there's no proxy or extra network hop, and a check takes under a millisecond. Prompts and data stay with you.

    in-process/no proxy

  6. 06

    Nothing breaks on day one

    Only critical policies block; high, medium and low ones warn. Start a new rule as a warning and make it critical once you trust it.

    shadow/suggest/enforce

Questions

Common questions.

01

What happens if the Kyvvu platform is down?

A running agent keeps deciding with the policies it already fetched. An agent that starts without any policies blocks its action steps by default; you can set it to allow them instead.

02

Does traffic go through Kyvvu?

No. There's no proxy and no extra network hop. The agent decides in-process; traces and incidents reach the platform afterwards.

03

Where do we start?

With the OWASP agentic default pack on one agent, in shadow mode: every verdict is recorded and nothing is blocked, so you see what would have been stopped on real traffic first.

Get started

Every agent has to ASK before it acts.

Start with one agent

See it on your own agents

A 30-minute demo with someone who builds Kyvvu: your agent, your rules, and the policy that decides each step.

Rehearse it

A 60-minute agent incident drill

Run the tabletop exercise with your team and get an after-action report with your gaps.

Want to see other teams and sectors? All use cases.