Next-generation agent platformthat actually gets work done.
huoku is an on-premise agent platform that turns scattered company knowledge into automated action — on your own infrastructure, with no external dependencies.
Knowledge that doesn't turn into action.
For most organizations, knowledge is scattered, isolated, and siloed. AI promises rapid change, but most platforms require moving your data outside your control.
Scattered knowledge
Critical information lives across dozens of systems — Confluence, SharePoint, file shares, intranets, email, ticketing tools. People spend hours hunting for things that already exist somewhere. No system connects the pieces.
Data sovereignty
Cloud AI platforms need access to your data. For many companies, sending sensitive business knowledge to an external vendor isn't an option — for regulatory, competitive, or policy reasons.
External dependencies
When you wire business-critical processes to a cloud AI vendor, you inherit a dependency. When their service goes down, your workflows stop. Operations should not run on someone else's infrastructure.
An on-premise agent platform that turns knowledge into action.
On-premise by default
huoku runs entirely on your own infrastructure. Data never leaves your network. Processes never depend on external APIs. Security is built in, not bolted on.
Agents that do the work
Agents don't just retrieve information — they call APIs, combine sources, and produce finished decisions. Knowledge becomes action.
Everything from one panel
Agents, identities, knowledge collections, data protection, monitoring, and traces. One view — for operators, security, and audit.
Agents in production.
Sensor data and manuals, together. Anomaly detected, the right team alerted, work order opened. Same platform, your own infrastructure.
Predictive maintenance.
A scheduled workflow continuously analyzes machine state. An anomaly is detected before anyone has time to ask — and the corrective action starts in the background.
Sensor data + manuals
The agent fuses OPC-UA streams, MES history, and PDF manuals into one context.
- OPC-UA sensor streams
- MES production history
- PDF maintenance manuals
Model diagnosis
Spots the anomaly, retrieves prior cases, and proposes root causes with source citations.
- vibration +38%
- temperature ↗ 87 °C
- ≈ manual §4.2
Route to the right team
Routing follows identity and shift — Teams, email, or a work order, directly.
- @shift-tech
- Line 03 — predictive
- open source →
Work order + instructions
Creates a maintenance order in the ERP, attaches step-by-step instructions, closes the loop.
- ✓ order opened
- ✓ instructions attached
- ✓ window 14:00
Procurement assistant.
The agent compares offers and contracts side-by-side — price, coverage, terms, and special clauses in a single view, with source documents in context.
- Side-by-side comparison table — rows extracted automatically
- Citations to the original PDFs always visible
- Differences that matter for the decision are surfaced
The control room.
Agents, workflows, identity, monitoring, and tracing — inside the same admin panel. One window for operators, security, and audit.
Manage AI agents centrally.
One view for all your AI agents. Each card shows the agent's role, available tools, and live status.
- 01Agent overviewCard view shows the agent's role, tool count, and status at a glance.
- 02Interface tagsColor-coded tags (chat, rest, kafka) immediately indicate which interfaces an agent supports.
- 03New agent creationAdd agents in a few clicks — no programming, no DevOps tickets.
Permissions follow the data — chunk to answer.
Security is not a separate layer. It's part of the architecture — every chunk, query, and answer carries an identity.
Chunk metadata
Every chunk preserves the original ACLs — groups, users, and visibility tags.
Data Guard
Regex-based PII masking before storage — personal data never enters the index.
Vector database
Hybrid search returns only the chunks the asker is authorized to see. The answer cannot leak data.
Tracing
Every query, retrieval, and answer is traceable down to the user and the source.
Sensitive identifiers are automatically masked in the document pipeline — GDPR and EU AI Act are satisfied without giving up the AI value.
Patient Matti Meikäläinen, SSN 120464-123A, arrived at the clinic on 2026-04-12.
Patient Matti Meikäläinen, SSN [REDACTED_FI_SSN], arrived at the clinic on 2026-04-12.
The result: the AI only answers from data the user is permitted to see — without unbundling or duplicating your existing access model.
Building real-time data platforms for operators taught us one thing: if it doesn't run reliably on the customer's own infrastructure, it doesn't run at all.
We built huoku on the same principle — on-premise · production-grade · no external dependencies
AI on your own infrastructure.
On your terms.
Tell us briefly about your environment and your goals — we'll come back within two business days with a concrete next step.