Website beta Evidence-backed public record

An AI agent you can configure, inspect, and control

Holaryn Agent is a proprietary, standalone, local-first AI agent for coding, automation, and everyday work. You choose the model connection; configured remote providers receive the information you send to them.

Run ledger Reviewable
  1. Model boundaryLocal or configured remote connection
  2. Bounded workPrompt, schedule, workflow, or tool plan
  3. Authority checkConsequential work can pause for approval
  4. EvidenceEvents, checkpoints, outcomes, and trace export

Evidence helps you inspect a run; it does not guarantee the result is correct.

Connection boundary

Work with your model and your boundaries

Holaryn Agent has no silent fallback model vendor. You configure the connection used for model-backed work.

Choose the connection

Use a configured local, remote, aggregator, subscription, or custom model endpoint. Availability, terms, and retention remain the provider's responsibility.

Meet the agent where the task is

Work through command, web, desktop, and selected service surfaces. A capability's release, platform, prerequisite, and maturity still determine where it applies.

Explore features and their evidence

Run record

Durable work, visible decisions

Runs can retain events, checkpoints, outcomes, cancellation, steering, and trace evidence so work can be inspected instead of disappearing into one response.

Inspect what happened

A recorded event or completed test is evidence for that run. It is not a guarantee that an answer or external result is correct.

Keep authority explicit

Policies, dry-run plans, held actions, and approvals help bound consequential work. Approval does not make an external action risk-free, and unattended work may pause.

Published builds

Choose a release channel

Release channels describe whole builds. They do not describe the maturity of every capability inside a build.

External downloads are not yet approved

The reviewed distribution gate is closed. No Nightly, Beta, or Stable binary is offered until the required terms and first-release approval are complete.

Interaction and trust

Designed for accessible operation

Semantic structure, keyboard operation, visible focus, announcements, reflow, zoom, and reduced motion guide the product and this website.

Evidence, with open coverage visible

Automated checks and manual procedures exist, but refreshed product coverage across JAWS, NVDA, VoiceOver, and Orca is incomplete. No formal product conformance claim is made.

Review accessibility evidence and limits

Local state, explicit network paths

Durable state is local by default. Configured remote providers and optional connectors receive the information sent to them and operate under their own terms.

Review security and privacy boundaries

Find the right reporting path

Use versioned documentation and configured product feedback for ordinary problems. Send vulnerability details through the private security route.

Find support for Holaryn Agent

Operating model

How Holaryn Agent works

Configure a boundary, ask for bounded work, then inspect outcomes and authority decisions in the documentation for the current Stable release.

  1. Configure a model connection

    Open Stable documentation for model and provider setup.

    Read model connection guidance
  2. Ask for or schedule bounded work

    Open Stable documentation for scheduled and automated work.

    Read scheduling and automation guidance
  3. Review outcomes and approvals

    Open Stable documentation for authority and approval controls.

    Read autonomy and approval guidance

Choose an exact record

Start with a published build and its matching guidance.