🌐 CZECHAI — MASTER ECOSYSTEM CATALOG

One connected technology ecosystem that brings together proprietary models, GPU infrastructure, agents, data, applications, products and specific workflows into solutions ready for real-world operation.

🧭 MASTER CATALOG STRUCTURE

I. Services, products and industry solutions 🧰 — what CzechAI offers and delivers to customers.

II. Technology, product and development areas 🚀 — where the solutions come from and where the ecosystem is heading.

III. Claude Arzenál 🧬 — a standalone internal catalog of agents, swarms, MCP, skills, plugins, commands and operating rules.

IV. Verified metrics, data and infrastructure 📊 — numerical data with the source of measurement.

V. Main public platforms and outputs 🔗 — a single hub.

VI. Master catalog rules 📏 — methodology for completeness, deduplication and updates.

🧰 I. SERVICES, PRODUCTS AND INDUSTRY SOLUTIONS

🧠 Proprietary LLM & AI lab

🖥️ Proprietary GPU and LLM infrastructure — we run self-hosted models, a model gateway, intelligent routing, load balancing and automatic failover between our own and external models, so the client gains performance, control over data and operational independence.

🧪 Model adaptation and fine-tuning — we adapt models to Czech, industry terminology, corporate documents, tool calling and the client's specific processes, so that answers and actions match the organisation's real practice.

📚 RAG and knowledge systems — we connect Qdrant, hybrid search, knowledge graphs, internal documents and verifiable sources into our own Czech knowledge layer cz-rag, which delivers traceable answers over up-to-date data.

🤖 Agents and orchestration — we build specialised agents, MCP tools, multi-agent workflows, long-term memory and automatic task handover between roles, so complex work runs like a coordinated digital team.

🗃️ Data intelligence & data pipelines

🕷️ Data collection and enrichment — we obtain data from websites, documents, registers and the client's internal systems and handle scraping, OCR, normalisation, deduplication, matching and continuous updates.

🏛️ Public and reference sources — we connect ARES, ČÚZK, RÚIAN, ISIR, auctions, official notice boards, case law and other verifiable sources into a single data layer usable for analysis and automation.

🗺️ Geodata and map applications — we create geocoding, risk and thematic layers, interactive maps, spatial analyses, 3D visualisations and location intelligence for decision-making about territory.

🔄 ETL, API and reporting — we turn raw data into databases, APIs, dashboards, XLSX/CSV exports, alerting and follow-up automations; this includes worldmonitor at monitor2.czechai.io for continuous global event monitoring.

🏠 Real estate, valuation and investment

📈 Price Map V26 — from an address or listing URL it estimates the value of a flat, house or plot of land, gives a price range, comparable offers and a confidence level; available at shadow.czechai.io.

💎 Below-market opportunities — automatically finds undervalued listings, direct owners, off-market opportunities and other investment signals that can easily be overlooked in ordinary advertising.

🔎 Reality Portal and Copilot — combines listings, market data, expected yields, risks and property screening into a single working environment; the main platform is EstateX at estatex.club.

🔨 Auctions and buyouts — tracks auctions, compares starting and final prices against the market and prepares materials for buyouts, investment assessment and fast decisions.

📡 Lead Hub — delivers turnkey investment leads: direct owners, long-listed and repeatedly listed offers, motivated sales, undervalued properties, auctions and insolvencies, with output to a spreadsheet, API or CRM.

🌍 Land development prediction platform — from zoning plans, ÚAP, ZÚR and ÚSES it evaluates the probability of future conversion of land into building plots; the solution is sold exclusively to a single client.

📱 Essty / EstateX mobile app — a real estate agent in your phone, connecting listings, valuations, clients, notes and voice control into a single mobile working tool.

🎮 HádejCenu and Hypotown — interactive applications for working with property prices and locations that educate users, collect relevant prospects and at the same time work as marketing tools.

🏘️ Housing stock management — a white-label application for homeowner associations and property managers connecting resident communication, billing, fault reporting and tradesperson dispatch.

⚖️ Legal AI & document intelligence

📖 AI Lawyer — performs legal research over regulations, case law and the organisation's internal knowledge and returns answers with traceable sources for subsequent expert review.

📄 Contracts and filings — analyses contracts, flags risky provisions, prepares redlines, templates, draft documents, procedural filings and a clear chronology of a dispute.

⚔️ Litigation support — helps sort evidence, build arguments, transcribe hearings and distinguish speakers, and quickly assess the latest developments in a case.

🛡️ Governance and compliance — supports the implementation of GDPR, the EU AI Act, NIS2 and internal rules, including checking legal citations against our own database and creating audit documentation.

🛡️ AML, OSINT, fraud & security

🏦 AML and sanctions screening — screens individuals, companies and properties against sanctions, PEP, insolvencies, enforcement proceedings, UBO and other registers; it contains 18 modules, a risk score, a status lifecycle and a structured risk report.

🔍 OSINT and due diligence — traces connections, the public digital footprint and verifiable information about a subject and turns them into a clear basis for commercial, security or legal investigation.

🚨 CzechAI Shield and fraud-guard — supports the detection of insurance and other fraud through case prioritisation, anomaly detection, network links and a calculated fraud score.

🔒 Security guardrails — protects AI operations by detecting sensitive data, defending against prompt injection, managing roles and access, logging and creating auditable security records.

🏦 Banks, finance and insurance

🏠 Collateral valuation — the V26 model prepares a market estimate, price range and comparative data for LTV, mortgage, credit decisions and collateral checks.

📋 Underwriting and claims handling — sorts cases, reads documentation via OCR, adds valuations and external context and prepares materials for faster decisions by the responsible officer.

💳 Loans and financing — covers credit scoring, affordability and mortgage calculations and follow-up financial workflows; our own platform HypoStars at hypostars.cz offers a calculator and a partner portal.

📉 Kronos financial forecasting — uses ML models for financial and market forecasts and provides an analytical basis, not individual investment advice.

🧑‍💼 Banker Copilot — works as an assistant directly in the CRM, searches product knowledge, calculates LTV and DSCR and prepares materials, summaries and client correspondence.

🏠 Collateral Intelligence — continuously monitors collateral value, revalues the entire portfolio and gives early warning of a drop in value, a change in risk or the need for further review.

📊 Insurance Knowledge Base — unifies policy terms, exclusions, limits and valuation tables into a single searchable knowledge base with AI search.

🚨 Anti-fraud engine and claims handling — detects anomalies, prioritises suspicious cases and supports claims handling by automatically reading, sorting and checking documents.

♻️ Reinsurance and catastrophe modelling — prepares materials for reinsurance and models accumulated risks, exposure concentrations and the possible impact of large-scale loss events.

🏛️ ČNB reporting and institution database — creates the data basis for regulatory reporting and maintains a structured register of banks, insurers and other financial institutions.

⚙️ Regulated deployment — we design on-premises or private cloud architecture with an emphasis on data protection, auditability, access control and secure integration into the institution's processes.

🏛️ Government · 16 AI solutions for public administration

🧾 Public data analytics — a set of 16 AI solutions over administrative, tax, property and open data that help institutions find connections and prepare materials faster.

📌 Official notice boards and opportunities — automatically collects, sorts and filters municipal ordinances, auctions, sales, grant opportunities and public tenders published by municipalities and institutions.

👮 Supporting analytical tools — the Police AI platform and other systems speed up work with information, transcripts, documents and connections, while always preserving the decision of the responsible person.

🧾 AI tax detective — cross-references data from the tax administration, the cadastre and public registers and looks for discrepancies or risk patterns usable as a basis for further inspection.

🛰️ Notice board monitoring — continuously tracks new postings across municipalities, regions, ministries and other institutions and delivers relevant changes as alerts.

🛡️ Cyber defence AI and transport infrastructure — supports attack detection and forensic analysis and at the same time makes it possible to predict traffic load and plan infrastructure investment better.

☁️ Gov Cloud — multi-tenant AI infrastructure for public institutions with separated data, an agent catalogue, permission management and a secure connection to state registers.

🎓 AI education for institutions — CzechAI Academy offers practical training, AI literacy, governance, security and the use of AI in line with the organisation's role and responsibility.

📣 Marketing, Facebook and reach

🎛️ Marketing Command Deck — manages a team of AI specialists for strategy, content creation, acquisition, analytics, distribution and marketing process automation.

👥 The largest Facebook community in the Czech Republic — 230 000+ members — we own the largest Czech Facebook group in the field, complemented by a network of 241 community channels, of which 92 cover investment, real estate and finance, enabling organic reach, audience activation and targeted content distribution.

📬 Rozesílka.eu — 37 million contacts — enables bulk e-mail campaigns with segmentation by industry, region and type of subject through our own mail server and campaign tool.

📈 Performance, content and distribution network — connects copywriting, SEO/GEO, paid media, social networks, the e-mail and SMS lifecycle, CRO and reporting into one measurable acquisition system.

✉️ Rozesílka.eu — our own platform for working with data, campaigns and AI communication, providing segmentation, bulk distribution, personalisation and response measurement.

🔁 CRM, leads and sales workflow — finds, enriches and qualifies leads, records them in the CRM, manages follow-ups and gives the sales team continuous reporting.

🎙️ Voice, image, video and documents

🗣️ Voice assistants — communicate in natural Czech, handle telephone and conversational scenarios, transcribe calls and distinguish individual speakers.

🧑‍💼 Avatars and video — create talking digital assistants, product and training videos and an automated video pipeline from script to finished content.

🖼️ Visual and 3D creation — includes generating images, product visuals, 3D elements, interactive maps and presentation materials for marketing and professional use.

📑 Documents and reports — automates OCR, the processing of PDFs, contracts, reports, flipbooks and data outputs as well as the preparation of finished client materials.

🖨️ Digital Print Shop — the tiskarna.digital platform turns digital outputs into physical products such as printed matter, custom printing, merchandising and bespoke presentation materials.

🌐 Applications, portals and integrations

💬 CzechAI Chat, Agent MAX, Aurum and darkstar — agentic chatboxes provide access to models, tools, research, documents and specialised assistants in one user environment.

🏘️ EstateX and industry portals — ready-made platforms for real estate, finance, education, marketing and client self-service, complemented by the Digital Marketplace at trziste.digital.

🔌 Custom integrations — we connect APIs, MCP servers, n8n, e-mail, CRM, databases, spreadsheets, documents and other client systems into one working workflow.

🛠️ Turnkey websites and SaaS — we deliver the whole product from the first working version to live operation, including frontend, backend, payments, administration, monitoring and further development.

⚙️ Infrastructure, operations and DevOps

🐳 Production operations — we manage servers, GPUs, Docker, PM2, databases, model services, monitoring and operational automation across 1 500 services, 273 containers and 6 servers.

☁️ Private cloud and on-premises — we design infrastructure according to data sensitivity, regulatory requirements, the performance needed and the way it integrates into the client's environment.

📡 Reliability and oversight — we provide availability monitoring, backups, centralised logging, alerting and operating procedures for fast incident resolution.

🚀 Fast delivery — we can prepare the first working MVP for a specific problem within weeks and then extend it according to measured results and users' real needs.

🎓 Education, strategy and transformation

🏫 CzechAI Academy — academy.czechai.io is our own certification and training programme for individuals, teams, companies and public institutions, focused on the practical and safe use of AI.

🧭 AI strategy — helps choose priorities, use cases, expected value, risks, data readiness and the way AI is governed, so that the investment leads to a measurable result.

👥 Adoption in teams — through practical workshops, internal assistants, guides and support we set up work so that employees genuinely adopt AI and use it every day.

📊 AI audit and roadmap — assesses existing solutions, data, processes and costs and turns the findings into specific priorities, measurable goals and a realistic deployment plan.

🔍 AI Rentgen — independently reviews third-party AI solutions, data and infrastructure and shows exactly what works, what is missing, where risks arise and what is being paid for needlessly.

🎓 L1–L5 certification — covers the path from AI literacy through practitioner and automation specialist to governance and industry specialisation; the corporate package includes the whole team and an internal AI directive.

🎤 Simulators and training — the sales simulator works with model customer types and the training assistant helps financial advisers safely practise real situations.

🏫 Educational AI for schools — the teacher's assistant prepares lessons, tests and assessment materials, while the learning AI communicates by voice and works with cited sources.

🧰 Sole traders and small businesses

📨 Job opportunity finder — continuously searches for public and commercial enquiries and opportunities matching the field, region and capacity of a specific business.

💵 Price spy — tracks competitors' prices in the area, compares offers and helps set pricing that is competitive and profitable at the same time.

🧾 Autonomous invoicing clerk — automatically issues invoices, checks due dates, flags unpaid documents and prepares or sends reminders.

⚖️ Pocket lawyer — helps with contracts, complaints, receivables and the basic legal situations a sole trader deals with in everyday business.

☎️ Voice receptionist — takes calls, answers common questions, books appointments and passes on messages, so the business does not lose customers even outside working hours.

💶 Grant radar — watches newly announced grants and subsidies, filters them by company profile and flags suitable opportunities and approaching deadlines.

⭐ Reputation manager — tracks reviews and ratings across platforms, flags new responses and helps build brand credibility systematically.

🏢 Corporate sector

🔎 Company Screening Platform — screens business partners, suppliers and counterparties against registers and open data and creates a single risk profile of the subject.

🔗 Supply Chain Risk Monitor — continuously tracks insolvencies, ownership changes, sanctions and other risk events in the supply chain and gives early warning about endangered suppliers.

🎯 M&A Target Finder — finds suitable acquisition targets by industry, location, financial indicators, ownership structure and the investor's strategic criteria.

🌱 ESG Compliance Platform — supports reporting under CSRD and ESRS, collects and consolidates ESG data and creates traceable materials for internal and external reporting.

🛒 Procurement Optimizer — analyses purchasing, compares prices, evaluates suppliers and looks for room for savings, order consolidation and better contract terms.

👤 HR Screening & Background Check — screens candidates only from lawfully usable public and register sources and prepares structured material for a responsible HR decision.

📊 Business Intelligence Hub and Knowledge Management — connects company data with an internal knowledge base and lets employees search, analyse and use information with AI.

🛒 E-commerce

💰 Competitive Pricing Engine — continuously compares prices across e-shops and provides the basis for automatic or managed dynamic pricing.

📉 Churn Prediction and customer segmentation — predicts customer churn, creates RFM segments and makes it possible to target retention and acquisition campaigns more precisely.

🎁 Product Recommendation AI — recommends relevant products to each customer based on their behaviour, purchase history, item similarity and current context.

🚨 Fraud Detection — looks for suspicious orders, unusual payment patterns and possible abuse of accounts, coupons or complaint processes.

📦 Inventory Optimizer and Demand Forecasting — forecasts demand, recommends optimal stock levels and reduces stock-outs, surpluses and needlessly tied-up capital.

🚚 Shipping Cost Optimizer — compares price, speed and quality of shipping across Česká pošta, PPL, DPD, Zásilkovna and other carriers and picks the most advantageous option.

🛍️ Abandoned Cart Recovery, Review & Reputation Manager and Marketing Attribution — brings customers back to abandoned carts, manages reviews and shows which marketing channels really lead to an order.

🚚 Logistics and transport

🗺️ Route Optimization Engine — optimizes delivery and collection routes using TSP/VRP algorithms over the precise RÚIAN address system and accounts for time, capacity and stop order.

🚛 Fleet Management and Carrier Performance — manages the vehicle fleet, tracks vehicle utilization and evaluates carriers by price, reliability, speed and claims.

🏭 Warehouse Picking Optimizer and Cross-Dock Optimizer — shortens movement in the warehouse, optimizes picking order and plans cross-docking of goods without unnecessary storage.

📐 Load Planning Optimizer — uses 3D bin packing to plan cargo layout so that vehicle capacity is used to the maximum while weight and operating limits are respected.

⏱️ Delivery Time Predictor — estimates the real delivery time from the route, traffic, carrier history, warehouse load and other current conditions.

❄️ Cold Chain Monitor — monitors temperature and transport conditions for sensitive goods and alerts on cold chain breaches or the risk of spoilage.

↩️ Returns Management — automates the recording, sorting, approval and evaluation of returns and helps uncover their most common causes and possible abuse.

🏭 Manufacturing

📈 Production Planning Optimizer — plans production by orders, capacity, material and deadlines and tracks OEE to reveal downtime, performance losses and bottlenecks.

🔍 Quality Control AI — uses computer vision to detect defects and links visual inspection with statistical process control (SPC).

🔧 Predictive Maintenance — analyzes data from IoT sensors and operating history and predicts failures before they cause downtime or a costly breakdown.

🌐 Digital Twin Simulator — creates a digital twin of the production line for safe testing of changes, capacities, scenarios and impacts before touching real operations.

🏷️ Traceability System — provides track & trace of materials, products and batches and delivers precise traceability during quality control or a recall.

⚡ Energy Consumption Optimizer and Yield Optimization — optimizes energy consumption and production yield by machine, shift, product and current operating conditions.

👷 Workforce Optimization and Continuous Improvement Tracker — plans shifts by required skills and performance and records improvement suggestions in Kaizen and Lean mode.

💼 Professional services

💹 Project Profitability Analyzer — tracks the actual costs, time, billing and margin of individual projects and flags in time those turning unprofitable.

📝 Proposal Generator AI and Document Automation — creates business proposals, contract annexes and recurring documents from verified templates and available client data.

🗓️ Resource Capacity Planner — plans team capacity by projects, roles, deadlines and workload and helps prevent both overload and idle staff.

🧾 Invoice Automation — links completed work with billing, prepares documents automatically and checks their dispatch, due date and payment.

📚 Knowledge Base AI — turns company documents, experience and methodologies into a central knowledge base with fast AI search and traceable sources.

😊 Client Satisfaction Tracker — continuously measures client satisfaction, NPS and feedback and flags relationships that need a fast response.

🤝 Client Relationship Manager and Business Development CRM — manages relationships, contacts, communication history, business opportunities and the whole pipeline from first outreach to closed deal.

🏥 Healthcare

🗓️ Patient Appointment Optimizer — optimizes patient scheduling, clinic utilization and waiting times and predicts the risk that a patient will not show up.

🎙️ Medical Transcription AI — transcribes medical dictations and conversations into structured documentation and supports ICD-10 code assignment for subsequent expert review.

🩺 Clinical Decision Support — searches expert sources using RAG and prepares relevant materials; it never replaces the clinical decision of a qualified physician.

📄 Insurance Claim Automation — automates the processing of health claims and documentation checks using the ICD-10 and CPT classifications.

⚠️ Patient Risk Stratification and Medication Adherence — sorts patients by risk level and helps track treatment adherence so care can be focused early on the most at-risk groups.

👩‍⚕️ Staff Scheduling Optimizer — schedules medical staff shifts by specialization, availability, legal limits and expected workload.

📊 Healthcare Supply Chain and Revenue Cycle Analytics — tracks supplies, consumption, billing and operating economics and looks for savings without limiting access to care.

🚀 II. TECHNOLOGY, PRODUCT AND DEVELOPMENT AREAS

This part keeps individual technology, product and development intentions as separate records. Each line represents a specific capability, development direction or operational area; the items are not merged into general summaries.

🖥️ GPU, servers and model infrastructure

🖥️ Expansion of our own GPU and server infrastructure in the Czech Republic — increasing compute capacity for AI inference, fine-tuning, machine learning, data computation, content generation and rendering.

🎛️ Acquisition and operation of high-performance GPUs — building our own hardware base for language, image, voice, analytical and multimodal models.

⚡ Scalable compute capacity — providing performance to internal projects and external customers according to current load.

🧮 GPU compute rental — commercial provision of secured compute power for inference, training, fine-tuning, computation and rendering.

🔌 Model APIs — making our own, open and external models available through a single secured interface.

🚦 Central model gateway — managing our own, open and external language models from a single operating point.

🧭 Intelligent task routing — automatic selection of the most suitable model by price, speed, data sensitivity, domain and task complexity.

⚖️ Load balancing between models and GPU servers — distributing traffic by available capacity, latency and current load.

🛟 Automatic failover — another model or server takes over the task on error, unavailability or an exceeded operating limit.

🇨🇿 Czech sovereign AI platform — our own and open models, operation in the Czech Republic and independence from a single foreign provider.

🏢 Private cloud — a separate environment designed around data sensitivity, required performance and the rules of a specific organization.

🏠 On-premise deployment — operation directly in the customer's infrastructure, where sensitive data need not leave the organization.

🔋 Energy optimization of GPU infrastructure — managing consumption, load, cooling and use of available energy when running compute nodes.

🧪 GPU and HPC test environment — research, development, benchmarks, pilot deployments and model validation on high-performance infrastructure.

🧠 Models, RAG and knowledge systems

🧠 Secure operation of our own language models — controlled model deployment under our own control, with the option to tune performance, data and security rules.

🌐 Secure operation of open language models — deploying open-source models without dependence on a single commercial API.

🧪 Model fine-tuning — further training of models on prepared data for a specific domain, organization or work task.

🇨🇿 Specializing models for Czech — improving work with the Czech language, inflection, meaning, style and local context.

⚖️ Legal and expert terminology — adapting models to Czech legislation, case law and the expert terminology of individual sectors.

📄 Adaptation to company documentation — tailoring answers to the customer's internal documents, methodologies, products and processes.

🛠️ Tool-calling and working with tools — developing models able to safely call APIs, applications, registers, databases and automation workflows.

📚 RAG systems — precise search and answering over extensive documents, databases and the organization's current knowledge.

🔎 Hybrid search — a combination of full-text, metadata, relational filters, vector similarity and domain scoring.

🕸️ Knowledge graphs — linking people, companies, documents, events, properties, legal relations and other entities.

✅ Verifiable answer sources — automatic citations, traceability of the information used and the option of subsequent expert review.

🧩 Czech knowledge layer cz-rag — our own layer for working with Czech documents, registers, law and domain data.

🗄️ Vector databases — storing and searching embeddings over case law, legislation, documents and other knowledge collections.

🧰 Data and knowledge technologies — PostgreSQL, Qdrant, Neo4j, Weaviate, ClickHouse, Redis, Kafka, MongoDB and other database, streaming and analytical layers.

🗃️ Data, registers, OCR, geodata and analytics

🕷️ Automated data collection — gathering information from websites, documents, public sources, registers and the client's systems.

🧹 Data cleaning and normalization — converting varied inputs into a single usable and checkable structure.

🧬 Deduplication and matching — removing duplicates and linking records belonging to the same person, company, document or object.

🕶️ Anonymization of sensitive data — protecting personal and confidential information during data processing, analysis and transfer.

🏷️ Data classification and enrichment — adding categories, relations, context, scores and other derived attributes.

🔄 Continuous updating and archiving — keeping a history of changes and regularly refreshing data from live sources.

⚡ Real-time data pipelines — continuous monitoring, validation, deduplication, archiving and updating of millions of records.

🏛️ ARES — company and economic data usable for background checks, contact enrichment, analytics and business processes.

🗺️ ČÚZK and cadastral data — parcels, title deeds, ownership relations and other materials for real estate and asset analysis.

📍 RÚIAN — a precise address and territorial system for geocoding, logistics, maps and spatial analysis.

⚠️ ISIR, auctions and official notice boards — insolvencies, forced sales, published documents and other events important for monitoring opportunities and risks.

⚖️ e-Sbírka and court databases — regulations, sections, case law and verifiable legal sources.

🛡️ Sanction and reference lists — sources for AML screening, PEP checks, relationship vetting and risk management.

🔤 OCR in more than 90 languages — converting scanned documents, photographs and forms into machine-processable text.

📄 Document parsers — extracting data, tables, attachments, metadata and the logical structure of documents.

🎙️ Recording transcription and diarization — converting audio into text and distinguishing individual speakers.

🔀 Statement comparison — finding matches, differences, contradictions and links across multiple transcripts or documents.

📑 Automated document outputs — creating PDF, DOCX, XLSX, reports, proposals, contract annexes and client materials.

🗺️ Geocoding and GIS — converting addresses and parcels into spatial data, working with layers and location intelligence.

🔥 Heatmaps, risk layers and 3D visualizations — clear display of prices, events, risks, infrastructure and territorial relations.

🌍 Global monitoring worldmonitor — continuous geopolitical, economic, security and event monitoring of public and expert sources.

🤖 Agents, MCP, integration and process automation

🤖 Autonomous AI agents — they use tools, carry out real tasks and trigger follow-up processes.

👨‍💻 Specialized domain agents — roles for development, law, finance, real estate, marketing, security, research, design, data and operations.

🤝 Domain AI swarms — coordinated teams of specialists assembled by task type and required outcome.

🧠 Long-term agent memory — retaining relevant context, decisions and work in progress.

🔁 Handing work over between agents — automatic continuation of a task by another role without losing essential context.

🎛️ Central orchestration — recognizing the task, splitting the work, selecting specialists, coordinating progress and checking the result.

🔌 MCP server development — standardized tool bridges for safe AI work with real systems.

🌐 API interface development — connecting models and agents to products, registers, client applications and external services.

🧩 Connector development — linking AI with e-mail, CRM, ERP, accounting, databases, documents, spreadsheets and web applications.

⚙️ Automation of real business processes — the path from receiving a request through finding and verifying information to a decision basis, a document, a system update or sending the output.

🗂️ Central catalogue of the working layer — a register of agents, skills, commands, tools, MCP servers and ready-made teams.

💬 Corporate AI assistants and copilots — CzechAI Chat, Agent MAX, Aurum AI and darkstar with access to models, internal knowledge, databases, tools and live sources.

🔗 Integration of n8n, CRM and follow-up workflows — running repeatable processes across applications without manual data re-entry.

🛡️ Security, audit, governance and regulated operation

🔐 Data and communication encryption — protecting information in transit, at rest and during processing.

🪪 Identity, role and permission management — users and agents access only permitted data, tools and actions.

🕶️ Sensitive data protection — detecting personal, confidential and regulated information before it is used further.

🧱 Separation of individual customers — isolation of data, context, model inputs and operating environments.

🚫 Protection against prompt injection — detecting and limiting attempts to change the system's rules or misuse connected tools.

🛡️ Security guardrails — technical and procedural boundaries for working with code, data, documents and critical systems.

🧾 Audit trail — traceable records of inputs, outputs, sources used, tools and actions performed.

📡 Logging and monitoring — tracking availability, performance, errors, changes and security events.

✅ Model testing and validation — checking the correctness, stability, safety and suitability of outputs for a specific use.

🔍 Control of incorrect and unsupported outputs — verifying claims against sources, databases and system rules.

⚖️ AI governance and regulatory compliance — EU AI Act, GDPR, NIS2, ISO/IEC 42001 and related internal obligations.

📋 AI system register and risk management — a catalogue of the solutions used, responsibilities, risks, controls and lifecycle.

📝 Corporate AI policy — rules for safe, responsible and auditable use of AI across the organization.

👤 Human in the loop — AI prepares and verifies the materials, but the final decision stays with the authorized person.

🏭 Products, platforms and industry systems

🏠 Reality Engine — continuous collection, consolidation and updating of property listings, auctions, insolvencies, debt enforcement proceedings, state offerings, official notice boards and cadastral data.

📈 Price Map V26 and proprietary valuation models — multi-level AI valuation of flats, houses, land, recreational and commercial properties.

📊 Comparative property analyses — price range, comparable listings, confidence level and an explanation of the data used.

💎 Finding undervalued properties — listing scoring, market comparison and automatic alerts on investment opportunities.

👤 Direct owners and off-market listings — finding opportunities outside standard real estate distribution.

⏳ Long-running and repeatedly advertised listings — tracking listing history, price changes and signals of a motivated seller.

🔨 Auctions, insolvencies, debt enforcement proceedings and state sales — monitoring forced and non-standard sale situations.

🗺️ Land development prediction — assessing possible changes in land use from zoning plans, ÚAP, ZÚR, ÚSES and other planning documents.

🏘️ EstateX — a platform for real estate agencies, brokers, investors and other professionals.

🔍 RealityScan — vetting, collection and analytical processing of property information.

🤖 Reality Copilot — an AI workspace for search, valuation, background checks, documents, clients and business decisions.

📱 Essty — a mobile real estate and personal AI assistant with voice control, valuation, clients and live data.

📡 Lead Hub — delivering direct owners, undervalued listings, auctions, insolvencies and other leads into a spreadsheet, API or CRM.

🐕 Property Watchdog — continuous monitoring of properties, price changes, listing status and new risk or business signals.

🎯 Opportunity Engine — automatic search, evaluation and prioritization of investment opportunities.

🏢 Applications for homeowner associations and property managers — resident communication, billing, fault reporting, fund management and tradesman dispatch.

🏦 HypoStars — mortgages, loans, bank comparison, creditworthiness calculation, scoring and partner cooperation.

🧮 Banking AI tools — collateral valuation, LTV and DSCR calculations, portfolio monitoring and loan documentation preparation.

📋 Insurance underwriting — case triage, reading documentation, adding external context and preparing materials for decisions.

🚨 Claims settlement and anti-fraud — automated document reading, anomaly detection, prioritization of suspicious cases and network connections.

♻️ Reinsurance and catastrophe modelling — management of accumulated risks, exposure concentrations and regulatory documentation.

💹 Financial analytics and predictive models — scoring, company analysis, market and financial forecasting and automated reporting.

⚖️ Legal AI — Czech legislation, case law, legal research, contracts, redlines, filings, dispute timelines and citation checking.

🎙️ Transcription of court hearings and interrogations — transcription, diarization, comparison of testimonies and preparation of analytical materials.

👮 Police AI — recording interrogations, finding contradictions, working with information and supporting analysis without replacing the responsible officer.

🧾 AI tax detective — cross-referencing tax, property, corporate, cadastral and other public data.

🏛️ Public administration systems — official notice boards, records management, forms, public procurement, property data and control processes.

☁️ Gov Cloud — secure multi-tenant infrastructure with an agent catalogue and connections to public administration systems and registers.

🛡️ AML platform — sanctions, PEPs, insolvencies, debt enforcement proceedings, ownership structures, UBO, financial data, risk score and structured reports.

🔍 OSINT and due diligence — public registers, digital footprint, entity links, domains, contacts, documents and international databases.

🚨 Anti-fraud and security systems — financial anomalies, suspicious transactions, service abuse, insurance fraud and risky connections.

🖥️ Cyber defence AI — detection of attacks and security events and support for forensic analysis.

📣 Marketing Command Deck — orchestration of specialists for strategy, content, acquisition, advertising, analytics and automation.

🎯 Lead and CRM systems — finding, enriching, segmenting, scoring, qualifying and automatically passing contacts into CRM.

📬 Rozesílka.eu — e-mail, SMS and multichannel campaigns, communication automation and response measurement.

🛒 E-commerce tools — dynamic pricing, product recommendations, inventory, demand forecasting, fraud detection, abandoned carts, reviews and attribution.

🚚 Logistics systems — route optimization, load planning, vehicle fleets, warehouses and delivery forecasting.

🏭 Manufacturing systems — capacity planning, quality control, computer vision, predictive maintenance, digital twins and traceability.

⚡ Energy systems — optimizing consumption in buildings, production plants, data centres and GPU infrastructure.

🏥 Healthcare applications — documentation transcription, capacity planning, working with specialist sources and support for qualified decision-making.

🧰 Tools for sole traders and small businesses — invoicing, receivables, orders, contracts, subsidies, reputation and a virtual receptionist.

🏢 Corporate systems — supplier vetting, supply-chain risk, ESG, procurement, HR and finding acquisition targets.

📊 Business Intelligence and Knowledge Management — corporate data layers, knowledge bases, AI search and management dashboards.

🎙️ Voice, multimedia, documents and application development

🗣️ Czech voice models — speech synthesis and speech recognition tailored to the Czech language.

☎️ Phone agents and virtual receptionists — taking calls, answering questions, booking appointments and passing on messages.

🎧 Customer support — conversational client service, request triage and escalation of complex cases.

🗣️ Voice control of applications — controlling functions and workflows with natural speech.

🖼️ Image and product visualization generation — marketing, presentation, product and technical image outputs.

🎬 Video, avatars and training content — talking assistants, an automated video pipeline and multimedia education.

🎵 Music and audio outputs — generating and processing music, voice and other audio material.

🧊 3D elements and interactive content — visualizations, models, maps, presentations and other multimodal outputs.

📚 Personalized books and materials — automated preparation of content tailored to a specific user or target group.

🌐 Turnkey web applications — frontend, backend, authentication, administration, payments, monitoring and long-term operation.

📱 Mobile applications — native or cross-platform products connected to AI, data, voice and client systems.

☁️ SaaS products and client portals — scalable applications, self-service, dashboards and industry platforms.

🏷️ White-label solutions — ready-made applications, agents, data platforms and industry systems operated under the customer's brand.

🛍️ Digital Marketplace — an offering of digital products, applications, services, data outputs and automated solutions.

🖨️ Digital Print Shop and AI Design Studio — personalized printing, books, promotional materials, custom printing, merch and print-on-demand production.

🎓 Research, education, audit and transformation

🔍 AI Rentgen — independent audit of third-party AI solutions, models, data, security, infrastructure and regulatory readiness.

🛡️ Security and technology audits — verification of architecture, configuration, dependencies, operations, models and data flows.

🧭 AI strategy — selection of priorities, use cases, expected value, risks, data readiness and governance approach.

🗺️ Transformation roadmaps — concrete steps, measurable goals, responsibilities and an AI rollout timeline.

🎓 CzechAI Academy — an education and certification program for individuals, teams, companies and public institutions.

🏭 Strojovna — hands-on work with tools, models, automation, RAG, agents and real-world domain tasks.

🪜 L1–L5 certification — AI Literacy, practical use of AI, automation, governance, security and domain specializations.

🎤 Simulators and training applications — tools for sales staff, financial advisors, employees, management and public institutions.

🧪 Research, development and pilot deployment — experiments, prototypes, testing on GPU and HPC, and transfer of the results into production.

🤝 Cooperation with universities and research institutions — joint development, testing, expert validation and transfer of results.

📑 Technical documentation and product certification — preparation of materials for operation, audit, handover, verification and formal certification.

🛡️ Protection of source code, brands and know-how — legal, technical and operational protection of the value created.

💼 Commercialization, distribution and operational scaling

📦 Licenses — commercial provision of software, models, data layers and specialized modules.

☁️ SaaS — delivery of ready-made services and applications on a subscription basis.

🔌 API services — paid access to models, data, computation, background checks, scoring and automated functions.

🏠 Private cloud and on-premises installations — standalone deployment for companies, banks, insurers, the state and regulated organizations.

⚡ AI computing power provision — inference, fine-tuning, computation, content generation and rendering for external customers.

📊 Commercial data outputs — analytical reports, monitoring, alerts, scoring and industry data services.

🤝 Partner and distribution network — cooperation with banks, insurers, real estate agencies, advisors, integrators and public institutions.

🇪🇺 European partnerships and consortia — internationalization, joint projects, research and commercial development across the European area.

👥 Team expansion — growth of development, security, research, operations and business capacity.

🐳 Container and DevOps infrastructure — secure deployment, updating and scaling of applications, data services and models.

📡 Continuous availability monitoring — checks on services, performance, errors, capacity and dependencies.

♻️ Automatic service recovery — restart, failover and restoration of operations without unnecessary manual intervention.

💾 Backup and recovery — protection of databases, configurations, documents, models and operational data.

🚨 Alerting and incident management — fast problem detection, escalation and documented restoration of operations.

🌍 A unified European CzechAI ecosystem — connecting Czech AI models, agents, data, applications, infrastructure, education and commercial services.

🧬 III. CLAUDE ARSENAL — AGENTS, SWARMS AND TOOLS

In this master document the Arzenál is deliberately kept separate from customer services. It describes the internal working layer: who does the work, how teams are assembled, which tools they use and which operating rules they follow.

🧬 Claude Stack · Arzenál

🧠 Claude orchestrator — the central control brain recognizes the assignment, splits it into steps, picks suitable specialists and assembles a team precisely for the given task.

🧑‍💻 Specialists for every field — ready-made agents cover development, data, law, finance, real estate, marketing, research, design, security and the day-to-day operation of the organization.

🔧 Control of connected tools — within the permissions granted, the Arzenál works with registers, documents, code, applications, databases, APIs and automation workflows, so it does not stop at answers but actually gets the work done.

🕸️ Swarms, memory and RAG — agent teams hand context over to each other, share a knowledge layer and can be switched from Claude to an in-house LLM as needed without changing the working process.

🧭 How the two catalogs complement each other

🌐 The complete solutions catalog — describes what CzechAI offers customers: products, applications, industry solutions, infrastructure, data and services.

🧬 This Arzenál catalog — describes which agents, teams, skills, tool connections and control mechanisms the working layer is built from.

📊 The Arzenál in numbers

🤖 749 agent records — the full inventory of agent definitions available on the workstation; not 749 unique personalities, but every recorded entry including duplicates and variants.

🧠 687 unique agents — the number of distinct agents after removing duplicates from the 749 recorded entries; measured 7 August 2026.

🗂️ 23 professional categories — agents are sorted by field and type of work, so a suitable specialist can be picked quickly or a whole team assembled.

🧩 26 documented swarms — the catalog sets out 9 permanent domain teams and 17 line-ups designed for a specific type of task.

🔌 46 MCP servers in the detailed section — tool bridges that let agents work with databases, the browser, e-mail, registers, servers and in-house services.

🛠️ 1 009 skills, of which 997 are unique — an extensive library of skills loaded only when they are needed for a specific job; 1 003 items have their own file SKILL.md with a description.

⌨️ 77 slash commands — ready-made workflows launched straight from the prompt with a slash, replacing the repeated manual entry of more complex instructions.

🔌 26 plugins — they extend the system with work on documents, development, data, design, finance, law, marketing, sales, GitHub, Stripe, Vercel and other areas.

🪝 3 automatic hooks — czechai-guardrail.js, gitnexus and herdr-agent-state.ps1 monitor changes, dependencies and the state of the agent process.

📏 1 global rule set — aurakit-security.md sets security guardrails that apply in every session, regardless of the project currently open.

🤖 Agents by profession

💻 Developers and programmers — 107 — they build frontends, backends, APIs, mobile and 3D applications, work with specific languages and frameworks, and handle architecture, refactoring and custom code implementation.

🏗️ DevOps and infrastructure — 55 — they manage Linux, servers, Docker, Kubernetes, CI/CD, networks, monitoring, incidents, scaling, performance and secure production operations.

🗄️ Databases and data engineering — 20 — they design schemas, SQL queries, data pipelines, migrations, indexes, storage and database optimization for reliable and fast access to data.

📊 Data science and AI/ML — 18 — they carry out analyses, statistical modeling, machine learning, predictions, experiments, model evaluation and data preparation for AI.

🎨 UI/UX designers — 17 — they design interfaces, user journeys, design systems, visual style, accessibility, prototypes and the usability of digital products.

✍️ Copywriters and content — 56 — they write websites, articles, e-mails, ads, scripts, documentation and expert content, and adjust the tone, structure, readability and persuasiveness of the text.

📈 Marketers and growth — 31 — they handle positioning, campaigns, SEO/GEO, paid advertising, analytics, CRO, lifecycle, social media, automation and measurable growth.

🤝 Sales — 17 — they prepare prospecting, lead qualification, offers, argumentation, follow-ups, sales processes, CRM and support for closing deals.

🎧 Customer support — 6 — they answer customers, triage requests, prepare knowledge-base answers, escalate problems and help manage support quality.

💰 Finance and accounting — 20 — they support financial analysis, budgets, reporting, valuation, investment research, accounting workflows, cash flow and the management of economic risks.

⚖️ Law and compliance — 34 — they analyze contracts, legal documents, GDPR, regulatory obligations, internal rules, risks and the alignment of processes with the required regulations.

🕵️ Security and OSINT — 28 — they carry out security audits, threat intelligence, penetration testing, background checks on people and companies, evidence collection and evaluation of the public digital footprint.

🔍 Research — 45 — they search for information from multiple sources, verify facts, work with citations, track competitors and trends, and turn findings into a usable synthesis.

📋 Project and product — 14 — they manage assignments, backlog, priorities, roadmap, specifications, stakeholders, risks and the handover of work between product, development and operations.

🏘️ Architects and review — 53 — they design the structure of solutions and independently check code, security, performance, consistency, scalability and the correctness of the resulting design.

🧪 QA and testing — 16 — they prepare and run automated and manual tests, look for regressions, verify system behavior and check quality before deployment.

🏠 Real estate and valuation — 6 — they search for properties and assess price, investment potential, comparable listings, legal status and other inputs for real estate decisions.

🎬 Video, audio and creative — 17 — they prepare scripts, visuals, editing, animation, sound, voice, 3D elements and other creative outputs for marketing, education and product communication.

🧑‍🏫 HR and education — 5 — they support recruitment, people management, training, the creation of study materials, skills development and the practical rollout of AI into teams and schools.

🌐 Web, CMS and integrations — 37 — they build and manage websites, CMS, e-commerce, API connections, automation and integrations between applications, databases and external services.

🏢 Strategy and leadership — 12 — they help with decision-making, planning, running the organization, business strategy, prioritizing initiatives and coordinating specialists.

📝 Documentation and knowledge — 4 — they create technical and process documentation, guides, knowledge bases, summaries and materials for handing over know-how.

🧩 Other and general-purpose agents — 90 — they cover cross-cutting roles, orchestration, critique, verification, context management and tasks that do not fall into a single field.

🧩 Permanent domain swarms

🗂️🤝 01-czechai — 24 agents — the in-house team for AML, banks, insurers, real estate, Czech law, OSINT and the day-to-day operation of our own services.

📣🤝 02-marketing — 21 agents — connects positioning, copywriting, campaigns, creative production, performance measurement, automation and the entire customer lifecycle.

🔬🤝 03-vyzkum — 18 agents — provides multi-source research, fact verification, citations, competitive analysis and trend tracking.

💻🤝 04-vyvoj — 25 agents — covers architecture, backend, frontend, programming languages, debugging, testing and application refactoring.

🧪🤝 05-kvalita — 17 agents — performs code review, tests, security checks, independent critical review and final verification of the output.

🎨🤝 06-design — 10 agents — handles UI, UX, user research, accessibility, 3D design and visual storytelling.

🖥️🤝 07-infra — 20 agents — manages servers, containers, networks, monitoring, incidents, reliability and operational security.

⚖️🤝 08-pravo — 6 agents — analyzes contracts, compliance, GDPR, regulatory obligations and the management of legal risks.

🤖🤝 09-ai — 9 agents — builds agent teams, prompts, MCP tools, orchestrations and repeatable AI workflows.

⚡ Task swarms

🕵️🤝 OSINT — full background check — lead osint-spy coordinates checks on an individual or company; specialists handle AML, law, search and fact-checking, and the reviewer Evidence Collector checks the evidence; the team uses 21 tool agents and the OSINT Swarm, OSINT API, OSINT Prověrka, Crucix and SpiderFoot services.

🏠🤝 Search swarm — real estate — led by offmarket-realitni-hunter; it brings in the agents off-market-undervalued-hunter, realestate-agent, Investment Researcher, url-context-validator and service-auditor in parallel, while Evidence Collector checks the accuracy and supporting evidence of the result.

🏡🤝 Real estate — property check — combines a real estate agent, a lawyer, AML, OSINT and investment research for due diligence on a property, its owner, title, price and related risks.

🕵️🤝 AML and sanctions — screens a person or company against sanctions, PEP, insolvency and other risk signals and has the results checked by an independent reviewer.

⚖️🤝 Law and contracts — analyses contracts, legal risks, GDPR and compliance with Czech legislation using legal, document and privacy specialists.

🏦🤝 Banks and mortgages — combines mortgage calculations, creditworthiness, comparison of banking products, insurance and an investment view into a single financial assessment.

💻🤝 Web and application development — leads a product from architecture through frontend, backend, TypeScript and UI to an independent review of the finished code.

🐞🤝 Debugging and bug hunting — moves from symptom to root cause, links debugging, tracing, build fixes and tests, and ends with verification of the fix.

🔒🤝 Security audit — checks configuration, dependencies, secrets in code, architecture and possible intrusion paths using auditors, pentesters and SecOps specialists.

🖥️🤝 Infrastructure and incidents — handles downed services, outages, full disks and other operational problems from diagnosis through incident coordination to recovery and verification of operation.

🔬🤝 Research and desk research — delivers multi-source research with search, claim verification, citations, knowledge synthesis and a final adversarial review.

📣🤝 Marketing campaign — leads an engagement from positioning through copy, creative and paid advertising to measurement and the mandatory critical review of the campaign.

📊🤝 Data and analytics — combines data science, analytics, research, SQL and database optimisation for a statistically and technically verified output.

🎨🤝 Design and accessibility — combines UI design, UX research, visual storytelling and an accessibility audit, with the result checked by an accessibility testing specialist.

🤖🤝 AI, agents and MCP — designs agent architecture, prompts, MCP tools, workflows and task allocation, and then has them reviewed by a red-team specialist.

📥🤝 Data collection and scraping — collects data from the web, checks that URLs are live, normalises results, adds context and has the evidence output independently verified.

📚🤝 Documentation and knowledge — creates documentation, guides, cited knowledge material and know-how handover materials, and subjects them to a final critique.

🔌 MCP servers — connecting agents to the real world

📦 15 MCP via `npx` — the tools are downloaded on demand from the npm registry and give agents specific external capabilities without a permanent local installation.

🖥️ 12 MCP via SSH to the server — the agent uses a secure tunnel to services running on a remote server and works with them as with available tools.

🌐 11 MCP via HTTP/SSE — remote services communicate over a web interface and continuous events, so APIs, registries and custom production systems can be connected.

🟢 7 local MCP processes — the tools run directly on the workstation and serve local data, files or applications.

🐍 1 local Python MCP — a standalone Python process provides a specialised capability through the standard agent interface.

🔌 Plugins, skills, commands and automation

🧩 26 plugins — the installed layer includes Notion, Claude HUD, management of CLAUDE.md, code review, Codex, Compound Engineering, Context7, customer support, two datasets, design, engineering, enterprise search, feature development, finance, frontend design, GitHub, legal, marketing, operations, product management, productivity, Pyright LSP, sales, Stripe and Vercel.

🛠️ 1 009 skills in the full catalogue — the alphabetical groups contain: 0–9 3, A 80, B 35, C 89, D 66, E 24, F 33, G 64, H 17, I 24, J 9, K 6, L 31, M 50, N 35, O 18, P 103, Q 14, R 61, S 143, T 39, U 13, V 15, W 24, X 4, Y 1 and Z 8.

⌨️ 77 slash commands — ready-made commands such as /analyze and others run predefined workflows for analysis, development, testing, documentation, agent management and operational work.

🪝 `czechai-guardrail.js` — a security hook automatically watches changes and enforces control guardrails when working with code and sensitive parts of the system.

🕸️ `gitnexus` — the hook links changes with repository context and helps agents understand relationships, impacts and dependencies in the code.

🧠 `herdr-agent-state.ps1` — the hook maintains and passes on the state of a multi-agent process so that individual roles work with a shared context and do not lose track of the task.

📏 `aurakit-security.md` — the global security rule is loaded in all sessions and protects the work regardless of the specific project.

🆕 Newly added agents

🆕 `czechai-change-reviewer` — independently reviews changes after code is edited or written and assesses correctness, security, conventions and adherence to AuraKit rules.

🆕 `czechai-team-lead` — coordinates several independent agent teammates using a shared task list and direct messages.

🆕 `mrkt-automation-architect` — designs marketing automation in Mautic, n8n, forms, landing pages, campaigns and dynamic content.

🆕 `mrkt-budget-optimizer` — allocates budget across channels according to expected ROI, saturation and the carryover effect of campaigns.

🆕 `mrkt-clv-analyst` — calculates customer lifetime value and predicts further purchase behaviour using BG/NBD and Gamma-Gamma models.

🆕 `mrkt-competitive-intel` — builds competitor profiles, finds market gaps, tracks share of voice and prepares material for brand differentiation.

🆕 `mrkt-copy-chief` — writes and rewrites conversion copy for websites, landing pages, ads and email campaigns.

🆕 `mrkt-council-critic` — performs an adversarial review of marketing output before a campaign launch or copy publication.

🆕 `mrkt-creative-studio` — designs ad creative, marketing graphics, visual concepts and prompts for AI generation.

🆕 `mrkt-cro-optimizer` — optimises conversions on websites, forms, sign-ups, paywalls and other decisive points of the customer journey.

🆕 `mrkt-email-copywriter` — prepares email sequences, newsletters, B2B cold outreach and subject lines.

🆕 `mrkt-growth-hacker` — designs growth loops, referral programmes, free tools, directory distribution and viral mechanisms.

🆕 `mrkt-influencer-manager` — runs influencer campaigns from creator discovery through contracts and execution to ROI evaluation.

🆕 `mrkt-lifecycle-manager` — builds email and SMS lifecycle, segmentation, lead scoring, retention programmes, anti-churn processes and dunning for failed payments.

🆕 `mrkt-market-researcher` — processes Voice of Customer research, reviews and interviews and turns real customer language into marketing material.

🆕 `mrkt-master` — acts as the main orchestrator of the marketing fleet for a launch, a complex campaign or an entire marketing strategy.

🆕 `mrkt-measurement-lab` — designs tracking in GA4 and GTM, A/B tests, attribution, experiments and a measurement plan including hard-to-measure dark social.

🆕 `mrkt-mmm-analyst` — uses a Bayesian Media Mix Model to calculate per-channel ROAS, adstock, saturation and uncertainty intervals.

🆕 `mrkt-paid-ads-commander` — runs Google, Meta and LinkedIn campaigns including account structure, targeting, bidding, pacing and measurement.

🆕 `mrkt-positioning-strateg` — defines positioning, the ideal customer profile, the product category and the brand narrative before content creation.

🆕 `mrkt-revops-conductor` — connects marketing and sales through lead lifecycle, scoring, routing, SLAs and consent governance under GDPR.

🆕 `mrkt-seo-geo` — increases organic visibility in classic search engines and the citability of content in AI answers and LLM systems.

🆕 `mrkt-social-video` — creates social calendars and ready-made scripts for Reels, TikTok, Shorts and other short videos.

🖥️ Workstation and server — what runs where

💻 Workstation — when the computer is on it provides 749 agents, 1 009 skills, 77 commands, 46 MCP servers and 26 installed plugins.

🌐 Server 179 — works independently even without the workstation and, as measured on 3 August 2026, contains 731 agents, 979 skills, 77 commands and 34 MCP servers.

🧩 Plugins only on the workstation — the plugin layer is not installed on server 179; remote operation uses server-side agents, skills, commands and MCP connections.

🖥️ Claude Code 2.1.220 on server 179 — this version forms the runtime environment of the remote Arzenál instance.

A10

🧱 Internal organisation of the Arzenál

🪪 Every agent has its own system ID — the record contains the professional category, the Czech name, a brief description of the capability and the system identifier used at actual launch.

🗂️ Categories are for orientation, IDs for launching — the professional breakdown makes selection easier, while the orchestrator and the swarms work with verified identifiers of specific agents, not merely with generic role names.

🤝 A swarm has clearly assigned responsibility — the team distinguishes a lead, parallel specialists and a reviewer; for selected teams it also records connected services, tool agents and operational endpoints.

🛠️ A skill is loaded only when used — the library can contain more than a thousand skills, because the individual guides do not take up working context before a specific agent actually needs them.

🔌 MCP is recorded by how it runs — the catalogue distinguishes tools launched via npx, an SSH tunnel, HTTP/SSE, a local process or a local Python service.

🖥️ Workstation and server have separate inventories — the Arzenál distinguishes precisely which agents, skills, commands, plugins and MCP the workstation has and which ones server 179 runs independently.

A11

🚨 Mandatory operating rules

👁️ Visual verification before marking something “done” — an HTTP 200 server response alone is not enough; the result must actually be displayed and checked, for example with a screenshot.

🌐 Scraping and OSINT only from servers 134/179 — public data collection and OSINT operations are not carried out directly from the workstation.

🔴 AML core and production website only after confirmation — the sensitive core and the live website are not touched without explicit consent.

🧠 Memory reflex — relevant memory is loaded at the start of the work and new findings and decisions are saved at the end.

A12

📝 Note on inventory accuracy

📅 Two revision dates — the summary inventory of unique items is marked on the website with the date 7 August 2026, while the capability overview, the operational status and some lower blocks use the date 3 August 2026.

⚠️ Differing totals on a single page — the top overview states 23 categories, 26 swarms and 44 MCP, the detailed sections state 26 swarms and 46 MCP, and older lower blocks still show 22 categories and 41 swarms; this document therefore uses the itemised 23 categories, 26 swarms and 46 MCP and explicitly preserves the discrepancy.

🧮 Sum of the professional cards — the visible counts in the 23 professional categories add up to 708 items, while the header records 749 entries and 687 unique agents; these values describe different layers of the inventory and are not to be confused.

Source: arzenal.czechai.io · live catalogue checked on 23 August 2026. In the master this part is kept as a separate, non-duplicate technical layer separated from customer services.

📊 IV. VERIFIED METRICS, DATA AND INFRASTRUCTURE

No.	Item	Value	Measurement source
M1	⏱️ Continuous development and operation	8 years (2018–2026)	brand history
M2	🧩 Own applications, modules and outputs	1 550+	conservative count; the public hub alone contains 6 521 unique URLs in 110 folders
M3	🌐 Live subdomains on czechai.io	130 (HTTP 2xx)	asset scan 2 Aug 2026
M4	🌍 Live hostnames across domains	153	asset scan 2 Aug 2026
M5	⚙️ Domains and subdomains in the nginx configuration	138	production server configuration
M6	🔗 Root domains held	16	export from registrars
M7	🖥️ Running PM2 production services	1 500 across six servers	`pm2 jlist` on each server
M8	🐳 Docker containers	273 running	`docker ps -q` on each server
M9	🤖 AI agents and specialized teams	4 020	in-house register of agents, swarms and skills
M10	🕸️ Of which in the production registry	73 in 40 swarms	`/v1/agents/registry/stats`
M11	📚 Of which the local definition library	749 agents + 1 475 skills	workstation inventory
M12	🛠️ Shared tools for agents	15	agent registry
M13	⚖️ Court decisions for Legal AI	578 032	`justice_decisions.db`, tabulka `decisions`
M14	📖 Legal regulations and sections	17 679 regulations / 30 712 sections	`czech_laws.db`
M15	🧠 Vector knowledge base — case law	617 362 vectors	Qdrant `judikatura_384`
M16	📚 Vector knowledge base — e-Sbírka	277 644 vectors	Qdrant `esbirka_384`; 23 collections in total
M17	🛡️ Sanctions records in the AML layer	3.6 million	AML platform; the live API currently holds 1 689 782 in memory
M18	🏢 Enriched ARES company records	3 255 092	`ares_enriched.db`, tabulka `companies`
M19	🏦 AML layer modules	18	AML platform
M20	📄 HTML pages in public production	4 586	`/var/www/router-static`, 25 GB
M21	🖥️ Own servers	6 + GPU server RTX PRO 6000, 96 GB VRAM	infrastructure
M22	📈 Apartment valuation accuracy — documented V24 backtest	R² 0.948	in-house V24 backtest; the current product line is V26

Note: the inventories were compiled in different layers and at different times. That is why, for example, the summary 4 020 agents, teams and skills, the local figure 749 agents + 1 475 skills and the Arzenál detail 749 records, 687 unique agents and 1 009 skills do not automatically rule each other out, but they must not be presented as the same metric without explanation. Before any public or investment use, each value should be re-confirmed against the stated source.

🔗 V. MAIN PUBLIC PLATFORMS AND OUTPUTS

🌐 CzechAI — https://czechai.io/

📘 Complete whitepaper — https://whitepaper.czechai.io/

🏭 CzechAI Strojovna — https://strojovna.czechai.io/

🎓 CzechAI Academy — https://academy.czechai.io/

🎬 Demos and implementations — https://show3.czechai.io/

🏠 EstateX — https://estatex.club/

🤖 EstateX Copilot — https://copilot.czechai.io/

📈 Price Map V26 — https://shadow.czechai.io/

🏦 HypoStars — https://hypostars.cz/

💬 CzechAI Chat — https://chat.czechai.io/

⚖️ AI Lawyer — https://law.czechai.io/

🧭 CzechAI Portal — https://portal.czechai.io/

🔍 AI Rentgen — https://rentgen.czechai.io/

🖨️ Digital Print Shop — https://tiskarna.digital/

🛒 Digital Marketplace — https://trziste.digital/

📬 Rozesílka.eu — https://rozesilka.eu/

🌍 Worldmonitor — https://monitor2.czechai.io/

🧬 Arzenál — https://arzenal.czechai.io/

📏 VI. RULES OF THIS MASTER

One piece of information, one primary place 📌 — services are in Part I, cross-cutting technologies and development in Part II, the internal agent inventory in Part III, metrics in Part IV and public links in Part V.

No loss of content ✅ — shorter and longer variants were merged into the most complete wording; only genuine or semantically identical repetitions were removed.

Preserved names and parameters 🏷️ — product names, domains, technologies, counts, revision dates, differing inventory totals and the notes on their methodology are all preserved.

Separating the offering from internal tooling 🧬 — Claude Arzenál is not presented as a standalone customer service; it is the working and tooling layer that delivers the services.

Verification before publication 🔍 — dynamic counts, operational status, regulatory claims and labels such as “largest” must be confirmed by a current measurement or a documented source before every public release.

✅ VII. COMPLETENESS CHECK

✅ Services and products — 19 numbered categories and 120 separate customer-facing items.

🚀 Technology and development — 9 separate technology categories and 161 detailed records, with no merging into short summaries.

🧬 Claude Arzenál — the Claude Stack introduction preserved, plus a full inventory of professions, standing swarms, task swarms, MCP servers, plugins, skills, commands, hooks, new agents, operational deployment and mandatory operating rules.

📊 Metrics — 22 numeric indicators including the source or measurement method and an explanation of the differing inventory totals.

🔗 Public outputs — 18 specific links to live products and platforms; technical router hubs and currently non-functional targets are not listed.

😀 Emoji check — every category heading and every recorded entry, service, agent item, metric, rule and public link has its own emoji.

🔎 Name check — an automated comparison against the main service catalogue and the separate Arzenál catalogue found no missing source name.

📌 Canonical purpose of this file: the single master catalogue of the CzechAI ecosystem. It removes the need to work in parallel with the compact service catalogue, the separate Arzenál catalogue and unsorted lists of technology or development areas.