Platform, integration and deployment.

Arkite is a manufacturing execution platform for manual operations. It turns manual assembly, kitting and inspection work into digitally structured workflows that are validated in real time and generate measurable production data. Arkite completes the digital manufacturing stack: the ERP decides what to build, the MES decides when to build, and Arkite tells operators how to build, step by step, validated and traceable. Arkite NV is headquartered in Genk, Belgium, with a second office in Aachen, Germany.

Arkite is a software platform. It is built around a central server, distributed workstation clients, a no-code workflow engine and open integrations with MES, ERP, PLC systems and smart tools. Projectors and sensors are enablers of the software, not the product itself. The same software core runs on every workstation, from a single pilot station to a multi-plant deployment, and all projects, users and production data are managed centrally on the Arkite server.

They are three levels of operational maturity on one software core, and each level includes everything in the level below. GUIDE provides digital work instructions with AR projection on the workbench; the operator confirms each step by touch screen, virtual button, hardware button or foot pedal. VALIDATE adds real-time 3D operator validation: the Arkite 3D sensor verifies every pick and assembly action and its sequence. VISION adds in-process product inspection with a high-definition Arkite Vision sensor that checks small parts, orientation and position during assembly.

Yes. Mixed-license environments are fully supported. Different workstations connected to the same Arkite server can run GUIDE, VALIDATE or VISION depending on the validation each task requires. A kitting station may run VALIDATE while a final assembly station with small, critical components runs VISION. Several license keys of different tiers can be combined on one server, and all workstations share the same projects, users and data infrastructure.

Most digital work instruction tools display information but do not validate execution. Arkite projects instructions directly on the physical workbench, validates each step in real time with a 3D sensor (VALIDATE), inspects product quality at component level during assembly (VISION) and records structured data for every action: cycle times, step durations, errors and torque values. It also integrates bidirectionally with MES, ERP, PLC systems and smart tools. The result is a documented, validated and data-generating production process.

No. Arkite is built on a no-code workflow engine. Process engineers define jobs, such as an assembly, a kitting task or an inspection, then teach the detections for picking bins, parts, tools and virtual buttons. Projections and text instructions are generated automatically from those detections. Engineers then structure the job in numbered steps and sub-assemblies, add text, images, video and icons, and add logic where needed: conditions, loops, variants and communication with external systems.

Variants are managed through jobs, job options and job variants. Job options are properties for which the steps differ, such as color, size or optional parts, and their combinations form the job variants. Shared steps are defined once, and variant-specific steps run only when they apply. The correct variant is selected on the operator screen, by scanning a barcode or automatically from the MES. Variants can be imported from and exported to CSV files, and there is no practical limit on the number of variants in a project.

Experience Levels adapt guidance to the proficiency of each operator. Three levels are available by default, Training, Basic and Advanced, and they can be renamed. Per level, engineers define which visualizations appear (for example extra text, images or video for new operators only), how projections behave, such as blinking during training, and how fast detections react on VALIDATE and VISION workstations, so experienced operators are not slowed down. The level is assigned per operator or selected at the workstation.

Yes. Instructions, button labels and all other text elements can be defined in as many languages as required. Instructions appear in the language of the logged-in user, and operators can switch language at the workstation at any time through the menu bar. Detection names can also be defined per language, so automatically generated instructions appear in the correct language. With the optional AI integration, texts can be translated automatically from the default language, per field or for an entire project.

Arkite integrates into the existing manufacturing ecosystem. It does not replace the MES or ERP; it connects them to the operator layer. Typical integrations are the MES for job orders, product variants and operator assignments, the ERP for scheduling and traceability data, the WMS for picking bin content, smart tools for torque control and feedback, PLC systems such as Siemens S7, BI platforms such as Power BI, and Active Directory for user authentication.

Communication modules are configured without programming. Supported protocols and devices include OPC UA, MQTT, Modbus TCP, Siemens S7, Open Protocol for smart tools, TCP/IP client and server, and HTTP web services. Database connections cover MySQL, MS SQL Server, Oracle, MariaDB and SQLite. Arkite also supports USB barcode and QR scanners, NFC readers and foot pedals, COM serial ports, pick-to-light, DMX lighting, Epson label printers, inter-workstation communication and PJLink projector control.

Jobs are created in the MES and pushed to the Arkite server through the API. Process engineers adapt the instructions in Arkite so they are practical for the physical workstation, and once approved, jobs are deployed to the correct workstations. When an operator logs in by badge scan or manual selection, the workstation retrieves the correct job and product variant from the MES automatically. This removes manual job and variant selection from the operator's tasks.

Yes. The Arkite API is a REST API for retrieving and modifying project data and for running operations on active workstations. Examples include scheduling tasks on a workstation when an order is received, loading a project on a specific workstation at the start of a shift, updating picking bin content from a WMS and reading the status of a detection. Full API documentation is published on SwaggerHub, an API tester is available in the software, and every user has a personal API key.

Yes. Smart tools connect through Open Protocol, Ethernet or Modbus. Arkite can enable the tool only during the correct step and, on VALIDATE and VISION workstations, only at the correct location on the workbench. It selects the required program or torque automatically, blocks the next step until the specified torque is confirmed and logs torque values per assembled unit. Small fasteners are validated by combining the pick from the correct bin, the tool position and the torque result.

Arkite uses a server-client architecture. The client software runs on each workstation, and the server stores all projects, users, resources and production records. The server is preferably installed on a separate physical server, virtual machine or VM in your private cloud. Minimum requirements are Windows Server 2012 or higher, or Windows 10 Enterprise LTSC 2019 or higher, 8 GB RAM up to 10 workstations and 16 GB above 10, and 100 GB SSD after OS installation plus 20 GB per connected workstation. Workstations connect over a reliable wired network, and the server's REST API on port 443 must be reachable by the workstations and by MES and ERP applications. The server backs up its databases every hour when changes are made; the frequency is configurable.

Every workstation feeds structured data back to the Arkite server: cycle times against expected cycle times per job and variant, step durations, errors with time and workstation, torque values from integrated smart tools, number of items picked per step, the operator and experience level used, and traceability records per serialized unit. Log data is stored in the Arkite logging database or in your own MySQL, MS SQL Server, Oracle or MariaDB database, with configurable retention.

The software includes four real-time dashboards: workstation status, job output per hour, day or week, cycle times per job, variant and workstation with trend line, and live step duration compared with the average. All production data can also be used in Power BI or another BI platform through a direct database connection, CSV export or the Arkite API. Manufacturers use this to identify bottlenecks, optimize cycle and takt times, and reduce scrap and rework.

Arkite records every operator action, validated step and deviation against a production order and serial number. In case of a quality issue, you can retrieve a complete log of every step performed on a specific unit, including the operator, the time, the result of each validation and the torque values applied. Product snapshots can be stored with the unit record, for example as an image in a database.

Arkite is used in aerospace, defense, automotive, medical device and general manufacturing, and several capabilities support compliance. Version management includes an optional approval workflow. Traceability is recorded per unit for steps, validations and torque values. Audit logging records every change to projects and settings with user, timestamp and previous and new values, in append-only records protected by a hash chain. Access is role-based, with Active Directory authentication.

All project and production data is stored on your own Arkite server, on-premises or in your private cloud, and remains fully owned by you. Log data can also be written to your own database. Workstations do not depend on an external cloud service during operation. Optional AI features used during engineering send text to the AI provider you select, only when enabled and configured with your own key.

Operator identification is optional. Workstations can run with a generic login, or operators can log in by badge or user account when individual traceability is required. Viewing, recording and taking snapshots of sensor streams is restricted to users with a dedicated role, and images are stored only when this is explicitly configured, for example product snapshots or recordings for engineering and support. Arkite NV processes personal data in line with the GDPR.

Users and roles are managed centrally on the server, so a user can log in on every connected workstation. Users can be local Arkite users or Active Directory users, with access granted per Active Directory group. Default roles include System Admin, User Manager, Editor, Approver, View Sensor Streams, View Dashboarding Tab and View Log Tables, and custom roles can be added. Operators have read-only access to projects in Production status.

Yes. Since software version 2026.2, engineers can generate jobs with AI. Based on a written description of the assembly, the software generates the job, job options, materials, detections and job steps. The engineer then reviews and finalizes the result before it is released to production. Detections are taught on the sensor image of the workstation, as with any other job.

Yes. With the optional AI integration, all instructions and text elements of a project are translated automatically from the default language: per field, for all empty fields or for the entire project at once. Instructions are shown in the language of the logged-in user, and operators can switch language at the workstation at any moment.

Since software version 2026.2, Arkite VISION includes an AI detection mode next to the existing detection modes. Teaching works as before, with one reference image and one expected result. AI detections are more tolerant when a part shifts or rotates slightly and transfer more reliably from one workstation to another after fine-tuning calibration, which simplifies rollout. They run locally on the workstation, are included in the VISION license and need no internet connection. Parts still need a defined position on the workbench.

AI job generation and translation use an integration with OpenAI or Anthropic, configured with the customer's own key. These integrations are optional, disabled by default and require an internet connection. Text is only sent to the selected provider when the integration is enabled, and it is processed under the customer's own agreement with that provider, which also determines whether data may be used for model training. Project and production data stay on your own Arkite server.

No. The AI detection algorithms of Arkite VISION run locally on the workstation and need no internet connection. Only the optional AI job generation and translation require an internet connection, and only when they are enabled.

The AI detection algorithms are included in the Arkite VISION license. AI job generation and translation are optional integrations in the Arkite software, configured with your own OpenAI or Anthropic key. See the Arkite AI page for an overview.