MyForge LabsHUNew project ↗Local AI / MyForge Labs
An assistant for internal knowledge and repetitive document work, in the environment you choose. We deploy on your infrastructure or operate an isolated private environment for your organisation.
An interactive story
Choose a question, then open the source behind the answer. See how scattered knowledge can become useful help.
Your knowledge
The documents, knowledge store and model sit within your company’s controlled environment. We agree who is responsible for operating them.
Approved documents move into an environment set up for you and operated by MyForge. We agree access, hosting and retention in advance.
The answer
Include the equipment ID, a short description of the problem and a photo. These give the service team a starting point for the next step.
Excerpt from the sample document
An issue report must include the equipment ID, a short description of the problem and a photo. The service team uses these to agree the next step.
The answer
Approved logos and templates are in the “Brand” folder on the shared drive. Ask your team lead for access if you cannot open it.
Excerpt from the sample document
Approved logos and templates are available in the Brand folder on the shared drive. Access is approved by the team lead.
The answer
The project owner checks the scope and delivery date. We send the version they have approved to the client.
Excerpt from the sample document
Before sending, the project owner checks the proposal’s scope and delivery date. The client receives only the approved version.
01 / The use case
We start with one clearly defined task, evaluated using your own examples.
Search procedures, manuals and product documentation with a checkable source alongside the answer. Staff access is limited to approved material.
Extract, summarise or classify items in quotes, messages and documents. Uncertain fields go for review; the designated staff member makes the final decision.
A draft response or data handover inside an existing business interface. We define what the assistant can read, what it can change and where human approval is required.
02 / Two operating models
The private environment can be yours or infrastructure we manage. Both options need explicit responsibilities and a defined data flow.
The model and knowledge store run on a server controlled by your organisation. We assess hardware, network and expected load, then prepare deployment and handover.
A good fit when your own environment is a requirement and an infrastructure owner is available. Hardware, updates, backups and support responsibilities are agreed in advance.
We operate an isolated private environment for your organisation under the agreed service scope. Your team uses the system through the interface we set up.
A good fit when you want a private environment without running a server yourselves. The proposal specifies hosting location and provider, access, support, backups and data export.
03 / Decision criteria
We evaluate Hungarian and domain-specific answer quality, document size and concurrent use together. These guide model, memory and any GPU requirements; we check the selected model’s commercial licence separately.
The model’s location is only one part: OCR, search indexes, backups, logs and every external connection also matter. We plan access and retention; external model calls can only be a separately agreed part of the solution.
Alongside hardware purchase or rental, we account for capacity, updates, monitoring and support. A new model or document source triggers a fresh evaluation of the affected tasks.
If a simple search tool or rule-based process meets the need, we recommend that. Local execution alone does not establish answer accuracy or data-protection compliance.
04 / Working together
One task, a few approved sample documents and expected answers. We clarify users, data types and the requirements of both operating options.
We evaluate answer quality, source references, response time and load using agreed questions. We also test denied access, missing answers and human handover.
You receive evaluation results, a data-flow description and an operating proposal. The pilot informs the deployment decision, including owners for updates, backups, recovery and support.
05 / Related development
Our document-processing prototypes and internal assistants show related approaches. Your data, hardware and users require their own pilot.
Related work
Next step
Tell us what it would work with, how many people would use it and whether it should run on your infrastructure or be managed by us. We can help assess the options.
Let’s discuss local AI ↗projects / atlas.workspace
Look behind the interfaces. Explore the solutions, inspect real screenshots and put two projects side by side.
9 projects
Choose up to two to compare.Client project
The complete system behind kisgepcentrum.hu: web shop, equipment rental, service intake, parts search and customer accounts — one platform shared with the in-store kiosk and staff workspaces.
Live business platform · kisgepcentrum.hu
Innovation prototype
An architectural plan-processing prototype: supported vector PDFs become editable quantities, cost estimates and professional exports.
Working prototype · private demo
Client project
A customer application for a tradesperson’s daily work: estimates, appointment requests, photos, messaging and service administration.
Public website and customer application
Club project
Digital tools for the pétanque community: tournament organisation, team score entry, spectator standings and shared sports information.
Public application
Client project
A custom bilingual salon website connecting its real space, stylists and services to a direct booking journey.
Public website
Client project
Developing an established clinic website: editable WordPress pages, two treatment areas and a coherent Hungarian–English patient journey.
Public website
Research collaboration
A workplace wellbeing platform that grew out of the instruments of a doctoral research project: research questionnaires, management reports and the Balance module, which rewards healthy habits.
Live platform · research collaboration with the University of Pécs Faculty of Health Sciences
In-house R&D
Automated processing of incoming supplier quotes, designed around a machine-parts distributor’s workflow: items extracted from e-mails, PDFs and spreadsheets feed a comparison matrix, with the original source behind every number.
Research prototype · synthetic data
Own product
A shared space for parents living apart: expenses, handovers, children’s requests and documents in one place, with two-sided approval.
Live · private, invitation-only family use
Try a different capability or clear the filters.
compare / projects
Compare the demonstrated capabilities and current status. Each project was built for a different purpose.
Select two projects from the cards above.
We understand the business problem, design the system and build it in steps you can inspect. Step inside the studio: choose a project and explore how research, human decisions and AI-assisted development connect.
Sources, open questions and an SSP brief.
Clarifying the goal and constraints together helps decide what belongs in the first usable version.
Before the next stageWe agree which problem to solve and how we will know it has been solved.
Research and sources
KGC-Bérlés
A single machine can be linked to a rental request, accessories, spare parts and a service case. The challenge was to preserve that context when a customer moves from the website to the store, or another team member takes over the conversation.
Planning and work breakdown
KGC-Bérlés
Customer-facing and internal web application development: catalogues, request submission, customer records, case-based communication, and counter and service views. The scope also includes the parts-engine interface, kiosk interaction, and editing, publishing and scheduling display programmes. In September 2026 the system gained the web shop, the full service workflow and customer accounts, and went into production.
Review and tests
KGC-Bérlés
Does the desktop parts diagram lead to the right part number and itemised enquiry?
The complete system behind kisgepcentrum.hu: web shop, equipment rental, service intake, parts search and customer accounts — one platform shared with the in-store kiosk and staff workspaces.
A single machine can be linked to a rental request, accessories, spare parts and a service case. The challenge was to preserve that context when a customer moves from the website to the store, or another team member takes over the conversation.
Customer-facing and internal web application development: catalogues, request submission, customer records, case-based communication, and counter and service views. The scope also includes the parts-engine interface, kiosk interaction, and editing, publishing and scheduling display programmes. In September 2026 the system gained the web shop, the full service workflow and customer accounts, and went into production. The platform has been running in production at kisgepcentrum.hu since September 2026, and development continues. Online rental requests require staff confirmation. Full ERP synchronisation, online card payments and automatic invoicing are separate workstreams; acceptance testing on the physical shop kiosk remains outstanding.
Suggested acceptance question · verification pending
Does the desktop parts diagram lead to the right part number and itemised enquiry?
The goal, context and acceptance criteria in one place.
# KGC-Bérlés / brief.md
> ILLUSTRATIVE SAMPLE · Based on public portfolio information. Not a client brief or approved project plan.
## Research · Gemini + NotebookLM
Reviewing the sources leads to a checkable brief. This example uses the portfolio and does not start live research.
## SSP / Goal
The complete system behind kisgepcentrum.hu: web shop, equipment rental, service intake, parts search and customer accounts — one platform shared with the in-store kiosk and staff workspaces.
## SSP / Context
A single machine can be linked to a rental request, accessories, spare parts and a service case. The challenge was to preserve that context when a customer moves from the website to the store, or another team member takes over the conversation.
## SSP / Constraints
Customer-facing and internal web application development: catalogues, request submission, customer records, case-based communication, and counter and service views. The scope also includes the parts-engine interface, kiosk interaction, and editing, publishing and scheduling display programmes. In September 2026 the system gained the web shop, the full service workflow and customer accounts, and went into production.
The platform has been running in production at kisgepcentrum.hu since September 2026, and development continues. Online rental requests require staff confirmation. Full ERP synchronisation, online card payments and automatic invoicing are separate workstreams; acceptance testing on the physical shop kiosk remains outstanding.
## SSP / Checkable outcome
Suggested acceptance question; verification pending:
- [ ] Does the desktop parts diagram lead to the right part number and itemised enquiry?
## Next handover · BMAD + Claude
- [ ] Separate sources, facts and open assumptions
- [ ] Review the SSP goal, constraints and acceptance criteria together
- [ ] Approve the implementation direction and verification approach
## Learning loop
Research may also start from previously approved lessons. A new finding needs a source, crystallization and an approved destination before propagation.
## Status and limits stated in the portfolio
The platform has been running in production at kisgepcentrum.hu since September 2026, and development continues. Online rental requests require staff confirmation. Full ERP synchronisation, online card payments and automatic invoicing are separate workstreams; acceptance testing on the physical shop kiosk remains outstanding.
Source: https://www.myforgelabs.com/en/case-studies/kgc-rental/Research can also use lessons that have already been reviewed and approved.
Record the actual result, environment and limits.
Select and check the lasting lesson within the raw experience.
How it becomes knowledgeAfter review, to an approved destination: a project document, decision or shared knowledge.
The approval pointBack to research, with better context.
Gemini + NotebookLMknowledge / connected.workspace
MyForge Superintelligence connects project context, specialised AI roles and developer decisions. A valuable lesson from one project becomes retrievable knowledge for the next.
01 / knowledge graph
Select a node for details. Drag to rearrange, or use arrow keys while focused. The list below also works on mobile. Zoom with Ctrl/⌘ + scroll, pinch, double-click, the − / + buttons, or the +, − and 0 keys while the graph is focused. Zooming in reveals the names, then the roles, then what each connection carries. Drag the background or use the arrow keys to pan.
Connections · what each one carries
context for the next taskbrief.mdresearch-notes.mdsources.mdagreed-brief.mdimplementation-plan.mdchange-set.diffreview-notes.mdverification.mdlearning-proposal.mdapproval.mdreviewed-knowledge.mdversioned Markdown sourcesRetrievable lessonsBrief, decisions, tasksReviewable implementationReal user journeysFeedback to developmentcapabilities / reusable methods
02 / discord handoffs
Every handoff has an owner, a purpose and a tangible output. This scripted example illustrates Discord-based collaboration: follow the request through research and development to human approval.
client-request.mdApproval continues only this local example.
MYFORGE / ERP In development · a closer look is coming soon
We have been developing an ERP that connects equipment rental, service, inventory and sales. Its purpose is to keep the history of an asset, partner or work order available at the next business step. A closer look at the development is coming soon.
Staff interfaces suited to each role
React + ViteRental · service · inventory · sales
Node.js + NestJSRelated data and queued background work
PostgreSQL + Prisma · Redis + BullMQShared foundation: roles, organisational separation and traceable operations.
BUSINESS MODULES / 01—05
The equipment journey from pricing through checkout and return to its related contract. Rental operations have their own business module.
Repair tasks, work orders and warranty cases are organised as connected workflows in dedicated service modules.
Modules cover products and stock movements from goods receipt and stocktaking to sales operations.
Partner records, business documents and related communication provide context for the same case.
Roles, audit trails and separated organisational workspaces define who can access each part of the business data.
ENGINEERING / STACK
A modular TypeScript monorepo with a Node.js and NestJS API, React–Vite workspaces, PostgreSQL–Prisma data handling and Redis–BullMQ background jobs. Dedicated access and tenant layers separate data belonging to different organisations.
The ERP in development is a separate project from the KGC-Bérlés web application. Modules and integrations are progressing incrementally; a public demonstration and operational rollout are separate delivery steps.
MYFORGE / WORKSPACE SETTINGS
Arrange this website around the way you like to explore. Changes appear immediately.
These settings apply only to this website. Layout preferences stay on this device; form entries are not saved here.
Workspace shortcuts do not run while you type in a form field.
❯ bmad Research · Gemini + NotebookLM
MyForge Workspace. Type help · Local navigation commands.
// Demonstration events.
Workspace opened. Choose a project or play the BMAD demonstration.
MYFORGE / WORKSPACE
Your open files stay right where you left them.
11.11@myforgelabs.com ↗