ESEN

Arcus — Control plane

Your agent fleet has already made more decisions than you can audit.

Without governance, AI scales chaos. Arcus measures and verifies each agent’s compliance — it doesn’t just declare it.

82% of organizations have AI agents operating in their infrastructure that are not inventoried — Cloud Security Alliance survey of 418 professionals, 2026.

Live decisions

Illustrative simulation

    The product

    Five layers. One control plane.

    Agnostic to LangGraph, AutoGen, CrewAI or whatever framework you already use.

    1. 01

      Visibility

      Which agents do I have?

      A live inventory and dependency map built from real traffic — not from what each team said it had. Detects shadow deps and what cascades if a resource fails.

    2. 02

      Control

      What can they do?

      Policies as code, fail-closed: if Arcus does not respond, the operation is blocked. Not an alert that arrives late — a guardrail.

    3. 03

      Observability

      What are they doing?

      Traces in OTel GenAI vocabulary, idempotent ingestion, cost and tokens under budget per agent and per organization.

    4. 04

      Trust

      Why should I trust them?

      Time-bound human approval and dual approvers on critical operations — separation between who drafts and who approves. The human stays in command.

    5. 05

      Assurance

      How do I prove compliance?

      Signed evidence, not screenshots. Risk classification with the cited article, incidents with their legal deadlines, reports exportable to PDF — mapped to the AI Act, ISO 42001 and NIST AI RMF.

    "Isolation at the database level — not just the application."

    The problem, in numbers

    The fleet grows faster than control.

    of organizations already use AI
    88%
    McKinsey, The state of AI in 2025 (1.993 organizaciones)
    are already experimenting with agents
    62%
    McKinsey, The state of AI in 2025 (1.993 organizaciones)
    have mature agentic AI governance
    21%
    Deloitte, State of Generative AI in the Enterprise 2026 (3.235 líderes)
    of projects will be cancelled by 2027
    40%+
    Gartner, comunicado de junio de 2025 (pronóstico)

    Compliance evidence that does not depend on who built your agents.

    Arcus records what every agent does, in any framework and on any cloud, and turns it into signed evidence you can present to your board or a regulator.

    Trust and compliance

    Compliance is not declared. It is measured.

    Secure

    Credentials and evidence live in a single place, designed for that risk.

    Assured

    Compliance is not declared, it is measured — coverage of each standard comes from signals Arcus records.

    Self-contained

    No dependency on a single vendor or cloud. It installs on top of what you already have.

    Auditable by design

    Full traceability, a test suite, deploys that abort if a migration fails. The product audits itself before it audits your agent fleet.

    in progress

    Isolation at the database level — not just the application.

    Certification process under way.

    An automatic check on every deployment blocks the build if that barrier is missing.

    Evidence, not screenshots.

    Every unit of evidence Arcus generates meets five conditions at once.

    • Operational originborn from telemetry, nobody writes it
    • Verifiable integritysigned, in an append-only log
    • Full contextagent, org, environment, timeline
    • Regulatory mappingAI Act, ISO 42001, NIST at once
    • Continuitya permanent flow, not last-minute collection

    Have 30 minutes to see this on your own fleet?

    FAQ

    • No, it supports it: it turns manual evidence collection into a continuous, auditable flow.

    A 30-minute conversation to find where your biggest governance risk is today.

    We reply within one business day. You will hear from someone on the founding team, not an SDR.