Controlled agent execution for teams that answer for every action.
hestiaOS classifies, evaluates and traces agent actions before execution — with public evidence for implemented controls, limitations and human approval points. We are selecting a small number of design partners in regulated EU domains to validate the Enterprise runtime against real workflows.
Explanatory mockup. Not a production deployment. Implementation status and limitations are documented in Public Evidence below.
hestiaOS is a pre-alpha governance-first runtime in active validation. Our public evidence surfaces document what is implemented today, what is experimental, and where human approval remains required.
Agent autonomy is arriving faster than the ability to account for it.
From August 2026, EU AI Act obligations for high-risk and general-purpose AI phase in. Regulated teams will need to show how automated actions are governed, logged and reviewable. hestiaOS is built to make that record a property of the runtime — not a report written afterwards. We support AI-Act-relevant traceability and transparency; we do not claim compliance on your behalf.
Why regulated teams talk to us
Three problems we hear from teams piloting agents in environments where actions have consequences.
“We can't reconstruct what the agent did.”
hestiaOS records input, actor, policy, decision, scope and timestamp as a causal trace — a replayable account of each governed action, not a log scraped together after the fact.
“Critical actions need a human in the loop.”
Actions are classified before execution. Sensitive or novel patterns are gated and fall through to human approval by design — the human gate is a runtime control, not a convention.
“Our data can't leave our environment.”
hestiaOS is designed for local-first, self-hosted deployment. The governance substrate runs where your data and your auditors already are — within a defined validation scope, not a cloud round-trip.
Where hestiaOS sits
Layer 7.5 — a deterministic governance substrate between probabilistic AI reasoning and real-world effects.
AI agent space
Probabilistic, untrusted, volatile. Model inference stays upstream.
Deterministic governance substrate
DSGK, CEG, CausalTraceGraph, Moirai. Governed, auditable, replayable.
Application space
Typed contracts, MCP tools, OS interfaces.
Public evidence
Verified results from the SPRIND Next Frontier AI submission. Each claim is backed by tests, benchmarks or code — with status and scope shown, not hidden.
Deterministic semantic governance
Intent lifecycle & execution gate
Non-gradient knowledge accumulation
Containerized reproducibility
Event-sourced vault · 10 invariants
Governed vs. ungoverned baseline
The Enterprise design-partner program
The Enterprise audit, compliance and operations layers are being built now. We build them with a handful of partners, against real workflows — so the first production-grade controls are shaped by teams who have to live with them.
What a design partner gets
- Direct line to the lead architect — your use case shapes the runtime roadmap.
- A scoped local-first deployment for one governed workflow you choose.
- Evidence pack: traces, policy decisions and limitations for your reviewers.
- Preferential terms when the Enterprise edition reaches general availability.
What we ask of you
- One concrete agent workflow where actions actually matter.
- A technical contact and a governance/compliance contact on your side.
- Structured feedback across the six-week validation window.
- Acceptance that this is pre-alpha: validation, not a production guarantee.
How the six weeks run
Scope & fit
Pick one workflow, define the governance questions, agree what evidence success produces.
Local deployment
Stand up hestiaOS in your environment. Wire policies, classification and the human gate.
Run & trace
Execute the workflow under governance. Capture traces, decisions and limitations.
Evidence review
Walk your reviewers through the record. Decide together on the path beyond the pilot.
One limitation, stated plainly
The Enterprise audit, compliance and operations modules are not yet production components. The governance core (DSGK, CEG, CausalTraceGraph) is implemented; the surrounding enterprise tooling is being built during this program.
This limitation is disclosed, not hidden. It is part of how hestiaOS is governed — and it is exactly why a design partner shapes what gets built.
Trust & limitations
Trust is produced by evidence, not adjectives. This documents current maturity. It is not a compliance claim.
✓ Implemented
DSGK constraint solver, CEG intent lifecycle (QUEUED → COMMITTED), CausalTraceGraph audit records, Moirai pattern-reuse engine.
⚡ Experimental
Enterprise audit/compliance/operations layer, full integration pipeline, production hardening — under active development.
👤 Human approval required
Critical actions are gated. Novel action patterns fall through to human review by design.
🔒 Security assumptions
Pre-alpha validation environment. Production security hardening is a roadmap deliverable, scoped with each partner.
🇪🇺 AI-Act relevance
We support traceability and transparency relevant to the AI Act. We do not claim the system is “AI Act compliant” on your behalf.
🤖 Content disclosure
Diagrams and the execution card on this page are explanatory mockups, clearly labelled, not production screenshots.
Who you work with
A founder-led team. In a design partnership you talk to the people building the runtime, not a sales layer.
Christian Walter
Owns the architecture, technology core and governance system. Leading the build of the SPRIND Stage 1 engineering team.
Christian Heilwagen
Runs delivery, roadmap and partner management. 14+ years in industrial electrical/automation systems, operational excellence and BI. Your interface for scoping and the pilot itself.
Planned Stage 1 expansion: 4 senior engineers + 1 ML researcher, an advisory board with frontier-lab expertise, and research collaboration with Forschungszentrum Jülich.
Questions reviewers ask
Is hestiaOS “AI Act compliant”?
What does “pre-alpha” mean for a pilot?
Where does our data go?
What does it cost to be a design partner?
Who actually does the integration work?
Validate governed agents before you have to defend them.
If your team is piloting agents where actions have consequences, talk to us. A short technical walkthrough is the fastest way to see whether a design partnership fits.
hestiaOS · pre-alpha governance-first runtime in active validation · SPRIND Next Frontier AI