# DDQ B.V. Autonomous AI and advanced technology, available now. DDQ helps companies and institutions automate work with autonomous AI, software and privately operated infrastructure. A small senior team designs, builds and operates secure solutions using local, external or hybrid AI according to the task. ## Main offers - [AI process automation with autonomous agents](https://ddq.nl/ai-automation/) (Nederlands: https://ddq.nl/nl/ai-automatisering/): Autonomous agents that take repetitive, browser-based and document-heavy work off a team, with clear permissions, logging and human control. For the right process this saves thousands of hours of labour. - [On-premise LLM and private RAG](https://ddq.nl/on-premise-llm/) (Nederlands: https://ddq.nl/nl/lokale-llm/): Open-weight language models on the customer's own hardware or DDQ servers in the Netherlands, connected to the organisation's own knowledge with RAG, so confidential data never has to leave. In use in medicine (Medical Copilot). - [Autonomous cloud](https://ddq.nl/autonomous-cloud/) (Nederlands: https://ddq.nl/nl/autonome-cloud/): A self-hosted, open-source European alternative to Microsoft 365 and Google Workspace: files, office, chat, email, code, wiki, single sign-on and private AI, on customer hardware or DDQ servers in the Netherlands. ## What DDQ offers now Companies and institutions lose substantial time and money to repetitive work, fragmented knowledge and manual processes. They need useful automation now, with an appropriate choice between local AI, external frontier models and deterministic software. Bring DDQ the goal, repetitive task, workflow, sensitive body of knowledge, existing system or difficult technical problem. DDQ selects and combines autonomous agents, local or external models, deterministic automation and conventional software, builds a working solution and can operate it on the customer's infrastructure or DDQ's privately operated infrastructure. ## Three ways to engage DDQ ### Build and operate digital products Applications, production backends and privately operated infrastructure, with sensing or dedicated hardware when the product requires it. Includes: Application development; Backend development and hosting; Autonomous and sovereign hosting; Scientific and integrated technology. ### Automate work with autonomous AI Autonomous agents and process automation using browser control, software tools, APIs, documents and organisational knowledge. The intelligence can be local, external or hybrid; RAG and fine-tuning are used where appropriate. Includes: Local and on-premise AI, RAG and fine-tuning; Autonomous AI and process automation; Agent frameworks and agent-ready services. ### Deliver research and funded innovation Open-science and EOSC implementation, citizen-science technology, technical proposal development, autonomous scoring and substantive delivery after an award. Includes: Open science and EOSC implementation; Grant and subsidy proposals. Full service and capability catalogue: - Application development: Mobile and web applications, from the first working version through store distribution, maintenance and long-term operation. DDQ has shipped medical, scientific, industrial and public-participation software since 2010. - Backend development and hosting: APIs, databases, data pipelines, identity, integrations and production backends, built and operated as one responsibility rather than handed between suppliers. - Autonomous and sovereign hosting: Private infrastructure on customer hardware or servers owned and operated by DDQ in the Netherlands, with open-source components, strong authentication and reduced hyperscaler dependence. - Local and on-premise AI, RAG and fine-tuning: Existing open-weight language models deployed on customer or DDQ infrastructure, connected to approved organisational knowledge with retrieval-augmented generation, evaluated and fine-tuned when adaptation is justified. DDQ does not train foundation models. - Autonomous AI and process automation: Agents and automations that interpret information, use browsers, operate software, call APIs, work with documents and complete bounded multi-step workflows. DDQ can use local models, external models from providers such as OpenAI or Anthropic, deterministic software, or a controlled hybrid. - Agent frameworks and agent-ready services: Provider-independent agent frameworks connected to approved tools and information, plus MCP, WebMCP, authentication and discovery layers that make websites and services usable by other agents. - Open science and EOSC implementation: Interoperable research services, FAIR data workflows, citizen-science platforms and practical European Open Science Cloud integration, backed by substantive H2020 COS4CLOUD and MONOCLE work. - Scientific and integrated technology: Software combined with sensors, signal processing, machine learning, mobile devices or dedicated hardware for scientific, medical and industrial use. - Grant and subsidy proposals: Call-fit assessment, technical concept development, TRL planning, proposal writing and autonomous scoring, followed by delivery of a substantive DDQ work package when the proposal succeeds. DDQ's range is a connected engineering capability, not a collection of separate departments. Customers enter through one of three broad needs. DDQ then combines only the capabilities required for the job and takes a defined scope that a three-person senior team can own. Larger programmes are delivered with established research, technical and consortium partners. ## How DDQ works Customers bring DDQ a subject, problem or opportunity. The team listens in depth, learns enough of the domain to become genuinely invested, and thinks alongside the customer about technology, regulation, implementation and funding. DDQ then builds the agreed work and remains loyal to the relationship and the system after delivery. 1. **Tell us what you are trying to achieve**: The customer brings the subject, problem or opportunity. A finished specification, chosen technology or matching service label is not required. 2. **We listen, learn and become involved**: DDQ talks with the customer in depth, learns the domain, asks technical and practical questions and starts thinking alongside the people who own the subject. 3. **Shape the technical and funding route together**: DDQ identifies the smallest serious first step, the wider opportunity, relevant constraints and partners, and whether the innovation and timing fit a European, national or regional grant. 4. **Agree a bounded first piece of work**: Before substantive research, architecture, proposal development or prototyping begins, DDQ and the customer agree scope, responsibilities, rights and a direct, co-funded or funded payment route. 5. **Build it and stay**: DDQ delivers the agreed system or work package, validates it with the customer and domain experts, and remains committed to the relationship through operation, maintenance or a responsible transfer. Commercial routes: - Direct assignment: The customer funds the scoped work directly when speed, certainty or programme fit makes that the right route. - Co-funded development: A grant covers part of the eligible project costs and the customer or consortium covers the balance. - Funded DDQ work package: In some programmes, all of DDQ's eligible work can be covered by the award. Other customer contributions, responsibilities or costs may still apply. Enthusiasm and an exploratory conversation do not create an unlimited free consultancy. When the work moves into substantive research, architecture, proposal development, prototyping or implementation, DDQ agrees a bounded scope, responsibilities, rights and a direct or grant-supported payment route. This protects the attention and loyalty that make the relationship valuable. Proposal collaboration: DDQ's work is often grant-friendly because it combines genuine technical uncertainty with a practical route from a lower technology-readiness level to a working, validated system. DDQ assesses whether customer or partnership development may be funded, shapes the technical concept and TRL progression, uses autonomous tooling to score drafts and find gaps, defines a credible work package and stays to deliver the funded work. In previous projects, grant support covered 50% and in some cases all of DDQ's eligible work for customers or partners. Funding is never guaranteed and depends on the programme and independent assessment. ### Best fit - Organisations spending too much skilled time on repetitive, browser-based, document-heavy or fragmented workflows - Companies that want to apply current AI privately and securely - Healthcare organisations working with sensitive data and consequential workflows - Research institutions that need scientific software, open-science services or participant-facing technology - Companies, universities and consortia preparing a serious research or innovation funding proposal - Organisations with technically unusual work spanning software, AI, sensing, mobile or hardware - Organisations anywhere that value privacy, security, interoperability, autonomy and direct access to senior builders ### Typical outcomes - A repetitive task or multi-step workflow automated with measurable time savings - A browser-using agent that operates approved software and services within defined permissions - A local or on-premise language model deployed with organisational knowledge through RAG - A model evaluation and fine-tuning route where adaptation is justified - A bounded autonomous agent connected to approved documents and tools - An agent-ready website, API or service with discovery and authentication - Automation integrated with existing workflows, applications and APIs - A maintained mobile application, backend or data product - A privately hosted production service on customer or DDQ infrastructure - An EOSC-compatible open-science or citizen-science service - An integrated sensor-to-software system - A technically credible proposal and a substantive DDQ work package ### Start with - The goal or persistent problem - The current workflow and affected users - Available knowledge, systems and data - Data sensitivity and required controls - Desired timescale ### Operating model and guardrails DDQ is deliberately small. Customers work directly with Norbert, Joep and Demelza. Decisions do not move through departments, and automation handles much of DDQ's administration and internal operations. This produces unusually fast execution and directs more of the customer's budget to senior technical work. DDQ does not claim the customer's medical, scientific, legal or operational authority. Qualified owners retain domain decisions. Scope, permissions, evidence, security, logging and human control are made explicit for each system. AI boundary: DDQ deploys and integrates existing models. It provides local and on-premise inference, RAG, evaluation and fine-tuning, but does not train foundation models. Hermes Agent is a provider-independent agent framework, not a model; it can work with locally hosted or external models. ## Origin and direction The name DDQ was originally imagined as an alternative to DDS, De Digitale Stad, one of the pioneering Dutch online communities. It later became a convenient, memorable digital address. In 2010, DDQ became a technology company during the first generation of native app development. The next generation of useful AI will not live only in software. DDQ is building toward systems that connect physical measurement, dedicated hardware, local inference, autonomous agents and sovereign infrastructure. The aim is technology that can observe, reason and act while remaining under the operator's control. DDQ became a B.V. in 2018 and is now a team of 3. Its work has progressed from native apps to sensing, machine learning and autonomous AI for medical, scientific and industrial applications. ## People You work directly with Norbert, Joep and Demelza. We are technically strong, supportive, enthusiastic and clear communicators who care deeply about the work itself. Automated overhead lets this small senior team move unusually quickly and put more of a customer's budget into the work. We never pretend to know a customer's field better than they do; we learn enough to think alongside them and bring an extra technical edge. We do cool things. DDQ travels the world to help set up scientific campaigns and spread the practice of crowd and citizen science, from the Arctic to Zimbabwe. We plan that travel to minimize its environmental impact wherever practical. ## Principles - We evaluate new technology early and use it where it has a practical advantage. - Science advances when knowledge, instruments and participation are open. - We prefer open source and open-weight models whenever possible. - Privacy and security are architectural requirements, especially in medicine. - Your data, infrastructure and technological future should remain under your control. - A small senior team, no bureaucracy, direct contact with the people who build. ## Capabilities GDPR/AVG-aware architecture and data minimisation. Information security for sensitive systems. EU AI Act-aware system design and AI governance. Standards-based and regulation-aware system design. European open-science infrastructure. Sensor devices and firmware. ML/AI systems. Computer vision. Data pipelines. Mobile and backend. On-premise deployment. Local AI inference. AI process automation. Browser use and browser automation. Provider-independent agent frameworks. Edge computing with ML on mobile devices. Open source and open weight models. Agentic AI. MLX contributions and maintenance. Native app development. Backend development and production hosting. Private LLM hosting. Retrieval-augmented generation (RAG). Language-model evaluation and fine-tuning. EOSC implementation and FAIR research services. Grant and subsidy proposal development. Agent discovery, MCP and WebMCP. Scientific and medical software. ## Current customers and work - Antoni van Leeuwenhoek: Medical technology - Maastricht UMC+: Medical technology - [NeuroMind Academy](https://neuromind.academy/): Medical Copilot and AI-supported medical education - Radboud University, Department of Astrophysics: Black Hole Finder ## Past public work - RIVM: iSPEX Polar Edition - Oxfam Novib: Applications for farmers and DataLab ## Research partners - Plymouth Marine Laboratory - Plankton Planet - Leiden University - ZMT Leibniz - IGB Leibniz - Institut de Ciències del Mar (ICM-CSIC), Barcelona ## Project collaborations - ODIX S.A., Belgium: Software and machine-learning technology for an innovative medical device. Supported through a STIPP subsidy. ## European research programmes - [COS4CLOUD](https://cordis.europa.eu/project/id/863463): Horizon 2020, grant 863463. DDQ developed MOBIS within the international research-infrastructure project, contributing software, backend and research work to interoperable citizen-science services connected with the European Open Science Cloud. - [MONOCLE](https://cordis.europa.eu/project/id/776480): Horizon 2020, grant 776480. DDQ developed mobile and low-cost sensing technology for water-quality observation and contributed to open system delivery, the iSPEX 2 processing chain, documentation and international field deployment. ## Operational proof - [Sovereign cloud infrastructure](https://cloud.ddq.nl/): DDQ owns and operates multiple servers for its sovereign cloud, including identity, collaboration, code, data and private-AI services. - [Dedicated AI infrastructure](https://cloud.ddq.nl/): Two dedicated AI machines serve DeepSeek Flash as a locally hosted model and run the provider-independent Hermes Agent framework continuously under DDQ's control. - [TOP / trombose.net](https://top.trombose.net/): A long-running medical production system that processes approximately one million patient files each year. ## Case studies ### [Medical infrastructure at production scale](https://top.trombose.net/) TOP / trombose.net is a long-running medical production system processing approximately one million patient files annually. Proof: Sixteen years of medical software experience, operational continuity and responsibility for sensitive production workloads. ### [Smartphones as scientific instruments](https://pocket.science/) DDQ helped turn smartphones into instruments for spectropolarimetry, light-pollution measurement, water observation and public participation in science. Proof: iSPEX, Dark Sky Meter, Mini Secchi, peer-reviewed work and international research partnerships. ### [Astronomy with people and machine learning](https://www.blackholefinder.org/) Black Hole Finder combines public participation, astronomical data and machine learning in ongoing work with Radboud University Astrophysics. Proof: A deployed platform, an active university relationship and independent coverage by NRC, Gizmodo and Space.com. ### [Sovereign AI in operation](https://cloud.ddq.nl/) DDQ runs private inference and persistent autonomous agents across owned servers and two dedicated AI machines. Proof: DeepSeek Flash provides local model inference. The provider-independent Hermes Agent framework supplies persistent autonomous tool use. Both operate alongside Authelia identity and production services on infrastructure controlled by DDQ. ### [Medical Copilot™](https://medicalcopilot.eu/) Private, on-premise AI developed for a hospital environment and now in use by NeuroMind Academy. Proof: Medical Copilot underwent a privacy audit by Maastricht University, and the word mark is registered in Class 10. ## DDQ Cloud services DDQ Cloud: Sovereign infrastructure on hardware owned and operated by DDQ. DDQ operates multiple servers and two dedicated AI machines on infrastructure under its own control in the Netherlands. The platform includes open-source services, Authelia identity, production workloads, DeepSeek Flash for local model inference and the provider-independent Hermes Agent framework for persistent autonomous work. Background and practical setup notes: https://noodpakket.tech/ - nextcloud (cloud.ddq.nl): files, talk, office - mattermost (mattermost.ddq.nl): team chat - chat (chat.ddq.nl): AI assistant, open source and private - webmail (webmail.ddq.nl): email - git (git.ddq.nl): code repositories - sso (auth.ddq.nl): single sign-on - wiki (wiki.ddq.nl): knowledge base For organisations that require privacy, security and control, DDQ designs and deploys sovereign infrastructure on their hardware or ours. ## Brands Pocket.science: Citizen science apps and sensors. https://pocket.science - iSPEX: Smartphone spectropolarimetry lineage for atmospheric and water-quality research. https://ispex.org/ In the press: "Citizen science: Amateur experts": Nature, 2013-04-10. https://www.nature.com/articles/nj7444-259a - Black Hole Finder: Help astronomers discover black holes. https://www.blackholefinder.org/ In the press: "De jacht op een zwart gat begint op je telefoon": NRC, 2026-04-30. https://www.nrc.nl/nieuws/2026/04/30/de-jacht-op-een-zwart-gat-begint-op-je-telefoon-a4925690 In the press: "Help Track Down Baby Black Holes Using This Citizen Science App": Gizmodo, 2024-08-08. https://gizmodo.com/help-track-down-baby-black-holes-using-this-citizen-science-app-2000487054 In the press: "Help scientists find new black holes with this free smartphone app": Space.com, 2024-08-07. https://www.space.com/black-hole-finder-smartphone-app In the press: "Black Hole Finder launch": IAU General Assembly Newspaper, 2024. https://www.iau.org/static/publications/ga_newspapers/2024/ga-2024-6.pdf - Dark Sky Meter & Loss of the Night: Light pollution monitoring. https://darkskymeter.com/ - Mini Secchi: Citizen science water quality monitoring. https://minisecchi.citizenscience.app/ - Kairos Coach: AI-powered fitness coach. https://kairos.coach ## Products - [TOP / trombose.net](https://top.trombose.net/): Long-running medical production software processing approximately one million patient files annually. - [Medical Copilot™](https://medicalcopilot.eu/): Private, on-premise AI developed for a hospital environment, privacy-audited by Maastricht University and now in use by NeuroMind Academy. Medical Copilot is a registered word mark in Class 10. - iSPEX®: Pioneering smartphone spectropolarimetry work with scientific origins at Leiden University and an international research network. ## Publications - [A universal smartphone add-on for portable spectroscopy and polarimetry: iSPEX 2](https://arxiv.org/abs/2006.01519): Co-authored by Norbert Schmidt, DDQ - [Standardized spectral and radiometric calibration of consumer cameras](https://arxiv.org/abs/1906.04155): Co-authored by Norbert Schmidt, DDQ - [Measuring night sky brightness: methods and challenges](https://arxiv.org/abs/1709.09558): Co-authored by Norbert Schmidt, DDQ ## Featured in Nature, NRC, Gizmodo, Space.com, IAU General Assembly Newspaper. ## For AI agents - MCP endpoint (Streamable HTTP, no auth): https://ddq.nl/mcp/ Tools: company_info, cloud_services, contact, submit_inquiry (sends a project inquiry to DDQ on behalf of a person). Resources: llms.txt, company.json, status.json, AGENTS.md, notes. Prompt: evaluate_fit. - Full text with all pages and notes: https://ddq.nl/llms-full.txt - Markdown: send `Accept: text/markdown` to any page, or append `index.md` to a directory URL - Structured data: https://ddq.nl/api/company.json - WebMCP manifest: https://ddq.nl/.well-known/webmcp.json - Agent guide: https://ddq.nl/AGENTS.md - OpenAPI: https://ddq.nl/openapi.json - MCP server card: https://ddq.nl/.well-known/mcp/server-card.json - Developer documentation: https://ddq.nl/developers/ ## Contact - Email: web@ddq.nl - Web: https://ddq.nl --- # ddq.nl agent guide DDQ B.V. welcomes AI agents. Public company information requires no authentication. Agents acting for a person can also send DDQ a project inquiry through MCP. ## Main offers - [AI process automation with autonomous agents](https://ddq.nl/ai-automation/): Autonomous agents that take repetitive, browser-based and document-heavy work off a team, with clear permissions, logging and human control. For the right process this saves thousands of hours of labour. - [On-premise LLM and private RAG](https://ddq.nl/on-premise-llm/): Open-weight language models on the customer's own hardware or DDQ servers in the Netherlands, connected to the organisation's own knowledge with RAG, so confidential data never has to leave. In use in medicine (Medical Copilot). - [Autonomous cloud](https://ddq.nl/autonomous-cloud/): A self-hosted, open-source European alternative to Microsoft 365 and Google Workspace: files, office, chat, email, code, wiki, single sign-on and private AI, on customer hardware or DDQ servers in the Netherlands. ## Commercial fit Bring DDQ the goal, repetitive task, workflow, sensitive body of knowledge, existing system or difficult technical problem. DDQ selects and combines autonomous agents, local or external models, deterministic automation and conventional software, builds a working solution and can operate it on the customer's infrastructure or DDQ's privately operated infrastructure. Best-fit customers: - Organisations spending too much skilled time on repetitive, browser-based, document-heavy or fragmented workflows - Companies that want to apply current AI privately and securely - Healthcare organisations working with sensitive data and consequential workflows - Research institutions that need scientific software, open-science services or participant-facing technology - Companies, universities and consortia preparing a serious research or innovation funding proposal - Organisations with technically unusual work spanning software, AI, sensing, mobile or hardware - Organisations anywhere that value privacy, security, interoperability, autonomy and direct access to senior builders Start an inquiry with: - The goal or persistent problem - The current workflow and affected users - Available knowledge, systems and data - Data sensitivity and required controls - Desired timescale Guardrail: DDQ does not claim the customer's medical, scientific, legal or operational authority. Qualified owners retain domain decisions. Scope, permissions, evidence, security, logging and human control are made explicit for each system. AI boundary: DDQ deploys and integrates existing models. It provides local and on-premise inference, RAG, evaluation and fine-tuning, but does not train foundation models. Hermes Agent is a provider-independent agent framework, not a model; it can work with locally hosted or external models. ## Connect - MCP (Streamable HTTP): https://ddq.nl/mcp/ - OpenAPI: https://ddq.nl/openapi.json - Company data: https://ddq.nl/api/company.json - LLM overview: https://ddq.nl/llms.txt - Full text (all pages and notes): https://ddq.nl/llms-full.txt - Service status: https://ddq.nl/status.json - Markdown: send `Accept: text/markdown` to any page, or append `index.md` to a directory URL. ## MCP tools - `company_info`: current offer, ideal customers, outcomes, operating model, guardrails and supporting evidence. - `cloud_services`: DDQ Cloud and private-AI services with current availability. - `contact`: public contact details and address. - `submit_inquiry`: send a project inquiry to DDQ. Requires `name`, `email`, `goal` and `contact_consent: true`; optional `organisation`, `workflow`, `systems_and_data`, `data_sensitivity`, `timescale`. Use it only when the person you act for has asked to contact DDQ and agreed to share these details. Limited to 5 inquiries per hour per client. A person at DDQ replies by email. ## MCP resources and prompts - Resources (`resources/list`, `resources/read`): `llms.txt`, `company.json`, `status.json`, `AGENTS.md` and every engineering note. - Prompt `evaluate_fit`: takes a `task` description and returns instructions for judging whether DDQ fits it. Send JSON-RPC 2.0 over POST to the MCP endpoint. Stateless discovery uses `2026-07-28`; legacy initialization supports `2024-11-05`, `2025-03-26`, `2025-06-18`, and `2025-11-25`. Examples: https://ddq.nl/developers/ ## Etiquette - Identify automated traffic with a descriptive User-Agent. - Cache stable public data and avoid unnecessary repeated requests. - Attribute factual summaries to DDQ B.V. and do not infer unsupported capabilities. - Never call `submit_inquiry` without the person's explicit consent, and never use it for unsolicited sales or bulk messages. --- About DDQ # We build the difficult thing and stay with it. DDQ is a three-person technology company that builds and operates autonomous AI, process automation, digital products, scientific systems and private infrastructure for organisations with difficult or sensitive work. Customers work directly with Norbert, Joep and Demelza. We listen until we understand the real problem, choose the technology that fits, build a working system and remain involved after delivery. We are small because that keeps decisions, responsibility and technical knowledge close to the work. ## What DDQ does ### Autonomous AI and process automation We automate repetitive, browser-based, document-heavy and fragmented work. Agents can interpret information, use browsers, operate software, call APIs, work with organisational knowledge and complete bounded multi-step workflows. The result is less manual work, fewer handovers and more time for people to use their judgment. ### Local and on-premise AI AI can run on customer hardware, within a customer-controlled environment or on infrastructure operated by DDQ. We provide private inference, RAG, model evaluation and fine-tuning. DDQ deploys and adapts existing models; it does not train foundation models. ### Applications, backends and infrastructure We build mobile and web applications, APIs, databases, integrations and production backends, then operate and maintain them when that is useful. Our own infrastructure supports identity, collaboration, code, data, applications and AI without making a hyperscale provider the only route. ### Scientific and integrated technology Some problems cross the boundary between software and the physical world. DDQ combines apps, sensors, signal processing, machine learning, calibration, edge computing and dedicated hardware for medical, scientific and industrial work. ### Open science and EOSC DDQ develops interoperable research services, FAIR data workflows, citizen-science technology and practical European Open Science Cloud integrations. Our experience comes from substantive delivery in COS4CLOUD, MONOCLE and other international research projects. ### Innovation funding We assess call fit, help shape the technical concept, define realistic TRL progression, write and score proposals, and deliver a real technical work package after an award. In earlier projects, grant support covered 50% and in some cases all of DDQ's eligible work. Funding is never guaranteed. ## What makes DDQ different ### One team across the complete system Software, AI, infrastructure, mobile, sensors and hardware do not have to be handed between unrelated suppliers. The people discussing the problem are the people responsible for building and operating the solution. ### Autonomy is built into the architecture We can use external models from providers such as OpenAI or Anthropic, locally hosted open-weight models, deterministic software, or a controlled combination. The model, agent framework, tools and deployment remain separate choices. Hermes Agent is an agent framework, not a model. ### Production experience where mistakes matter [TOP / trombose.net](https://top.trombose.net/) processes approximately one million patient files each year. [Medical Copilot™](https://medicalcopilot.eu/) was developed for a hospital environment, underwent a privacy audit by Maastricht University and is now used by [NeuroMind Academy](https://neuromind.academy/). ### Emerging technology without a large-company delay DDQ has moved from early native apps to smartphone sensing, machine learning, private AI and autonomous agents since 2010. Much of our own operational and administrative overhead is automated, allowing a small senior team to move quickly. ### We understand funded innovation We know how to move technology from a lower readiness level toward a working, validated system. Our proposal tooling tests call fit, scores drafts and exposes missing evidence, while the team remains responsible for the actual technical work. ## Who works with DDQ - SMEs that want repetitive work automated without building an internal AI team. - Healthcare organisations working with sensitive information and consequential workflows. - Research institutions that need scientific software, open-science services or participant-facing technology. - Universities, companies and consortia preparing serious research and innovation proposals. - Organisations with technically unusual work spanning software, AI, sensing, mobile or hardware. - Organisations anywhere that value privacy, security, interoperability and operational control. ## The team behind DDQ [image: Norbert Schmidt] **[Norbert Schmidt](https://norbertschmidt.nl/)**Founder and owner [image: Joep van der Heiden] **Joep van der Heiden**Developer and researcher [image: Demelza Ramakers] **Demelza Ramakers**Project planning and office Norbert founded DDQ in 2010 and connects product direction, software, AI, scientific measurement and infrastructure. Joep builds software and contributes research expertise. Demelza keeps planning, administration and project work moving. We are technically strong, supportive communicators who become genuinely interested in the customer's subject. The name DDQ was first imagined as an alternative to DDS, De Digitale Stad, one of the pioneering Dutch online communities. The company started during the first generation of native app development and became a B.V. in 2018. ## How DDQ works 1. **Start with the goal.** Email or call us with the problem, workflow or opportunity. A finished specification is not required. 2. **Talk directly with the team.** There is no sales handoff. We ask questions, learn the domain and identify what is actually worth building. 3. **Choose the smallest serious first step.** We define scope, responsibilities, data controls, evidence, timing and whether a funding route is realistic. 4. **Build and validate.** We implement the work with the customer and the people who hold the relevant domain authority. 5. **Operate, maintain or transfer.** We can continue operating the system on customer or DDQ infrastructure, maintain it, or arrange a responsible handover. ## Key facts **Company name** DDQ B.V. **Type** Independent technology company **Founded** 2010 **Legal form** Dutch B.V. since 2018 **Founder** [Norbert Schmidt](https://norbertschmidt.nl/) **Team** Norbert Schmidt, Joep van der Heiden and Demelza Ramakers **Headquarters** Kloosterweg 1, 6412 CN Heerlen, the Netherlands **Geographic scope** Worldwide **Website** [ddq.nl](https://ddq.nl/) **Core offering** Autonomous AI, process automation, software, private infrastructure and scientific technology **AI deployment** Local, on-premise, external or hybrid **Pricing** Scoped per engagement; direct, co-funded and grant-funded routes are possible **Contract terms** Scope, responsibilities, rights, controls, support and payment route are agreed for each assignment **Services** Autonomous AI, process and browser automation, local and on-premise AI, RAG, fine-tuning, applications, backends, private infrastructure, scientific technology, EOSC implementation and grant proposals **Communication** [web@ddq.nl](mailto:web@ddq.nl) and [+31 45 203 1008](tel:+31452031008); customers work directly with the team **Notable organisations** Antoni van Leeuwenhoek, Maastricht UMC+, NeuroMind Academy, Radboud University, RIVM, Oxfam Novib, Leiden University and Plymouth Marine Laboratory **Customers served** SMEs, healthcare organisations, research institutions, universities and international consortia **Projects delivered** Selected work includes TOP / trombose.net, Medical Copilot, COS4CLOUD, MONOCLE, iSPEX, Black Hole Finder, Dark Sky Meter and Mini Secchi **Operational proof** Approximately one million patient files processed annually; multiple owned servers; two dedicated AI machines **Competitors** The alternatives vary by assignment and include large system integrators, conventional app agencies, generic cloud AI platforms, specialist research platforms and internal development teams **Social** [Norbert Schmidt](https://norbertschmidt.nl/), founder ## Frequently asked questions ### Does DDQ train AI models? No. DDQ deploys and integrates existing models. We provide local inference, RAG, evaluation and fine-tuning when useful, but we do not train foundation models. ### Can DDQ automate browser-based work? Yes. Agents can use browsers, software interfaces, APIs, documents and approved organisational knowledge. The system is bounded by explicit permissions, logging, failure handling and human approval appropriate to the task. ### Does everything have to run locally? No. Local or on-premise AI is available when privacy, cost, resilience or control requires it. We can also use external models or a hybrid architecture when that is the better engineering choice. ### Can DDQ work with sensitive information? Yes, subject to a properly scoped architecture and agreement. Our experience includes medical production systems and private AI, but the customer retains medical, legal, scientific and operational authority. ### Can the work be funded through a grant? Sometimes. DDQ can assess fit, help shape and score the proposal, and deliver a substantive technical work package. Funding depends on the programme and independent assessment, so it is never promised. ### Does DDQ work internationally? Yes. DDQ works with customers and research partners internationally and has supported scientific campaigns from the Arctic to Zimbabwe. EU regulation and EOSC are areas of expertise, not geographic limits. ### How do we start? Send the goal, current workflow, affected users, available systems or data, sensitivity level and desired timing to [web@ddq.nl](mailto:web@ddq.nl). Do not send patient data, passwords or other secrets in the first message. --- Source: https://ddq.nl/about/ Site index for agents: https://ddq.nl/llms.txt --- Contact # Bring us the goal or the difficult problem. Contact DDQ when your organisation needs private AI, a digital product, scientific technology, privately operated infrastructure or a serious technical work package for a research and innovation proposal. **Email:** [web@ddq.nl](mailto:web@ddq.nl) **Telephone:** [+31 45 203 1008](tel:+31452031008) **Address:** DDQ B.V. Kloosterweg 1 6412 CN Heerlen The Netherlands ## What to include You do not need a finished specification or a chosen technology. A useful first message explains what should become possible, how the work is done now, who is affected and why the present situation is limiting. Where relevant, tell us which systems, knowledge or data already exist, how sensitive they are, what controls matter and the timescale you are working toward. - The goal, opportunity or persistent problem. - The current workflow and the people who use it. - Available data, knowledge, devices and existing systems. - Privacy, security, regulatory or deployment constraints. - An upcoming funding call, deadline or desired delivery window. We will first determine whether the work fits DDQ and whether a direct, co-funded or grant-funded route is realistic. Substantive research, architecture, proposal development or prototyping begins only after scope, responsibilities, rights and payment have been agreed. ## Existing services and technical contacts Existing DDQ Cloud users can sign in through the [DDQ Cloud login](https://auth.ddq.nl/). Developers and automated agents can use the public interfaces described in the [developer documentation](https://ddq.nl/developers/). To report a security issue, email [web@ddq.nl](mailto:web@ddq.nl); the canonical reporting details and expiry date are published in [security.txt](https://ddq.nl/.well-known/security.txt). Do not send passwords, private keys, patient data or other unnecessary sensitive material in an initial message. ## How contact data is used Information sent by email or telephone is used to understand and answer the inquiry, take requested pre-contractual steps, maintain an existing working relationship and meet applicable legal obligations. DDQ does not use a public contact form on this site. AI agents acting for a person can send an inquiry through the `submit_inquiry` tool of the [MCP server](https://ddq.nl/developers/#inquiry); it reaches the team as an email. More detail about website requests, correspondence, retention and privacy rights is available in the [privacy notice](https://ddq.nl/privacy/). --- Source: https://ddq.nl/contact/ Site index for agents: https://ddq.nl/llms.txt --- Developers and agents # Public interfaces for agents. DDQ publishes a small JSON API and an MCP endpoint for current company information, service availability and contact details. One MCP tool, `submit_inquiry`, sends DDQ a project inquiry on behalf of a consenting person. Nothing on this page creates an order or changes business data. ## Quick reference | Interface | Endpoint | Format | | --- | --- | --- | | Company data | [`GET /api/company.json`](https://ddq.nl/api/company.json) | JSON | | Service status | [`GET /status.json`](https://ddq.nl/status.json) | JSON | | MCP | [`POST /mcp/`](https://ddq.nl/mcp/) | JSON-RPC 2.0 over Streamable HTTP | | OpenAPI | [`GET /openapi.json`](https://ddq.nl/openapi.json) | OpenAPI 3.1 JSON | | MCP server card | [`GET /.well-known/mcp/server-card.json`](https://ddq.nl/.well-known/mcp/server-card.json) | JSON | | Full text | [`GET /llms-full.txt`](https://ddq.nl/llms-full.txt) | Markdown: every page and note | | Any page as Markdown | `Accept: text/markdown` | Markdown | | AI Catalog | [`GET /.well-known/ai-catalog.json`](https://ddq.nl/.well-known/ai-catalog.json) | AI Catalog 1.0 JSON | ## Authentication The endpoints listed above contain public information and require no API key, account, cookie or OAuth token. Do not send credentials. DDQ Cloud applications and customer systems are separate services with their own access controls; publication of these read-only interfaces does not grant access to them. ## JSON API The company endpoint is the canonical machine-readable representation of DDQ’s public company record. Request it with an identifying user agent and cache stable data instead of polling it unnecessarily. ``` curl --fail-with-body \ -H 'Accept: application/json' \ -H 'User-Agent: ExampleAgent/1.0 (ops@example.org)' \ https://ddq.nl/api/company.json ``` Successful responses use `application/json`. Failures anywhere below `/api/` use RFC 9457 problem details with `type`, `title`, `status`, `detail` and `instance`. Clients should branch on the HTTP status and `type`, not on the prose in `detail`. ``` HTTP/1.1 404 Not Found Content-Type: application/problem+json { "type": "https://ddq.nl/developers/#problem-not-found", "title": "Not found", "status": 404, "detail": "No API resource exists at this path.", "instance": "/api/missing" } ``` ## Rate limits Requests to `/api/company.json` and `/status.json` share a limit of 100 requests per client address in each fixed one-hour window. Every API response includes the current `RateLimit-Policy` and `RateLimit` structured fields plus `RateLimit-Limit`, `RateLimit-Remaining` and `RateLimit-Reset` for older clients. Reset values are seconds until the current window ends. ``` RateLimit-Policy: "hourly";q=100;w=3600 RateLimit: "hourly";r=99;t=1800 RateLimit-Limit: 100 RateLimit-Remaining: 99 RateLimit-Reset: 1800 ``` After the quota is exhausted, the API returns `429` problem details and a `Retry-After` header. Respect that value before retrying. The MCP endpoint has no numerical quota for reading, but abusive or unsafe traffic may be restricted. `submit_inquiry` accepts 5 inquiries per client address per hour. Cache `tools/list` using the hints returned by the current protocol. ## MCP negotiation The MCP server is stateless. Apart from `submit_inquiry`, every tool is read-only. It supports current MCP `2026-07-28` discovery and the legacy initialize handshake for `2024-11-05`, `2025-03-26`, `2025-06-18` and `2025-11-25`. Current requests mirror the method in `Mcp-Method`, carry the protocol version in the HTTP header and request `_meta`, and use `Mcp-Name` when a method addresses a named tool. ### Current stateless discovery ``` curl https://ddq.nl/mcp/ \ -H 'Content-Type: application/json' \ -H 'Accept: application/json, text/event-stream' \ -H 'MCP-Protocol-Version: 2026-07-28' \ -H 'Mcp-Method: server/discover' \ --data '{ "jsonrpc":"2.0", "id":"discover-1", "method":"server/discover", "params":{"_meta":{ "io.modelcontextprotocol/protocolVersion":"2026-07-28", "io.modelcontextprotocol/clientInfo":{"name":"example-client","version":"1.0.0"}, "io.modelcontextprotocol/clientCapabilities":{} }} }' ``` ### List capabilities without invoking a tool ``` curl https://ddq.nl/mcp/ \ -H 'Content-Type: application/json' \ -H 'Accept: application/json, text/event-stream' \ -H 'MCP-Protocol-Version: 2026-07-28' \ -H 'Mcp-Method: tools/list' \ --data '{ "jsonrpc":"2.0", "id":"tools-1", "method":"tools/list", "params":{"_meta":{ "io.modelcontextprotocol/protocolVersion":"2026-07-28", "io.modelcontextprotocol/clientInfo":{"name":"example-client","version":"1.0.0"}, "io.modelcontextprotocol/clientCapabilities":{} }} }' ``` The deterministic list exposes `company_info`, `cloud_services`, `contact` and `submit_inquiry`. The first three take no arguments and are annotated `readOnlyHint: true`. Listing these definitions does not execute a tool. ### Resources and prompts `resources/list` returns `llms.txt`, `company.json`, `status.json`, `AGENTS.md` and every engineering note; read one with `resources/read` and its `uri`. `prompts/get` with name `evaluate_fit` and a `task` argument returns structured instructions for judging whether DDQ fits that task. ### Sending an inquiry Call `submit_inquiry` only when the person you act for has asked to contact DDQ and agreed to share their details. `name`, `email`, `goal` and `contact_consent: true` are required; `organisation`, `workflow`, `systems_and_data`, `data_sensitivity` and `timescale` are optional. The team receives the inquiry by email and replies to the given address. Identical inquiries are sent once and return the same reference, so retries are safe. Validation problems return a tool result with `isError: true` and an explanation. ``` curl https://ddq.nl/mcp/ \ -H 'Content-Type: application/json' \ --data '{ "jsonrpc":"2.0", "id":2, "method":"tools/call", "params":{"name":"submit_inquiry","arguments":{ "name":"Jane Doe", "email":"jane@example.org", "organisation":"Example Hospital", "goal":"Reduce the time staff spend matching referral letters to patient records.", "data_sensitivity":"Medical data; must stay on-premise.", "contact_consent":true }} }' ``` ### Legacy initialize ``` curl https://ddq.nl/mcp/ \ -H 'Content-Type: application/json' \ -H 'Accept: application/json, text/event-stream' \ --data '{ "jsonrpc":"2.0", "id":1, "method":"initialize", "params":{ "protocolVersion":"2025-11-25", "capabilities":{}, "clientInfo":{"name":"example-client","version":"1.0.0"} } }' ``` After initialization, a legacy client sends `notifications/initialized` and may request `tools/list`. The server does not issue a session identifier because none of its public tools requires connection state. ## MCP errors MCP failures remain JSON-RPC responses. Invalid JSON uses `-32700`, invalid request envelopes use `-32600`, unknown methods use `-32601`, and invalid parameters or unknown tool and prompt names use `-32602`, and unknown resource URIs use `-32002`. Current stateless requests also use `-32020` for request/header mismatches and `-32022` for unsupported protocol versions. Clients should preserve the request `id` when correlating responses. ### API problem: not found The request path does not identify a published API resource. The response status is `404`. ### API problem: method not allowed The resource exists, but the request method is not supported. The response status is `405` and the `Allow` header lists supported methods. ### API problem: rate limit exceeded The client address has exhausted its fixed hourly quota. The response status is `429`; wait for the number of seconds in `Retry-After`. ### API problem: internal error The server could not produce the requested representation. The response status is `500`; retry cautiously or contact DDQ if the condition persists. ## Discovery and change policy - [AGENTS.md](https://ddq.nl/AGENTS.md) gives concise guidance for automated visitors. - [llms.txt](https://ddq.nl/llms.txt) and [index.md](https://ddq.nl/index.md) provide readable company context; [llms-full.txt](https://ddq.nl/llms-full.txt) adds every page and note. - Every page has a Markdown version, returned for `Accept: text/markdown` and linked with `rel="alternate"`. - [The API catalog](https://ddq.nl/.well-known/api-catalog), [AI Catalog](https://ddq.nl/.well-known/ai-catalog.json), [agent card](https://ddq.nl/.well-known/agent-card.json) and [WebMCP manifest](https://ddq.nl/.well-known/webmcp.json) link the machine interfaces. The public interfaces are versioned conservatively, but consumers should tolerate new object properties. Material changes will be reflected in the OpenAPI document, discovery metadata and this page. Questions can be sent to [web@ddq.nl](mailto:web@ddq.nl). --- Source: https://ddq.nl/developers/ Site index for agents: https://ddq.nl/llms.txt --- Privacy notice # Clear, limited use of personal data. This notice explains how DDQ B.V. handles personal data when someone visits ddq.nl, uses its public machine-readable interfaces or contacts DDQ. It does not replace the specific privacy and data-processing arrangements agreed for a customer project or operated service. **Controller:** DDQ B.V., Kloosterweg 1, 6412 CN Heerlen, the Netherlands. **Privacy contact:** [web@ddq.nl](mailto:web@ddq.nl) · [+31 45 203 1008](tel:+31452031008) **Last updated:** 2 September 2026. ## Data handled on this public site The public website has no advertising technology, behavioural analytics or public contact form. DDQ does not set tracking cookies on these pages. Like an ordinary web server, the infrastructure necessarily receives technical request data such as an IP address, date and time, requested path, response status, referrer when supplied, and browser or agent user-agent. DDQ uses that information to deliver requests, diagnose faults, protect the service, investigate abuse and understand aggregate operational load. Requests to the public JSON API (`/api/company.json` and `/status.json`) are limited to 100 per client address per hour. A one-way representation of the client address and the current request count may be kept for the active rate-limit window. Ordinary security and server logs can be retained for longer when reasonably necessary to operate and protect the service, investigate an incident or meet a legal obligation. Access is restricted to people who need it for those purposes. ## When you contact DDQ If you email or telephone DDQ, we handle the contact details and content you provide, together with relevant follow-up correspondence. This may include your name, organisation, role, email address, telephone number and information about the proposed or existing work. We use it to answer the inquiry, evaluate fit, take requested steps before a contract, carry out a working relationship, keep appropriate business records and comply with applicable obligations. Please do not include patient information, passwords, private keys or other unnecessary sensitive data in an initial inquiry. An AI agent acting for you can also send an inquiry through the `submit_inquiry` tool of DDQ's MCP server. DDQ then receives the details the agent submits (your name, email address, organisation and project description) by email, together with the agent's self-reported client name and user agent. A copy is stored on DDQ's own server, with a one-way hash of the sending network address used only to limit abuse. These inquiries are handled in the same way as an email inquiry. ## Purposes and legal bases Depending on the context, DDQ processes these limited data because this is necessary to take steps at your request before entering into a contract, to perform a contract, to meet a legal obligation, or for legitimate interests in communicating with organisations and securely operating the website and business. Where a different basis is required, DDQ will explain it in the relevant context. DDQ does not sell personal data and does not use public-site request data for personalised advertising. ## Sharing, infrastructure and transfers DDQ operates this website on infrastructure under its control. Technical providers may still transmit or process limited data where this is necessary for internet connectivity, email delivery, security or professional services. Information may also be shared when required by law or needed to establish, exercise or defend legal claims. DDQ seeks appropriate contractual and technical safeguards when another party processes personal data on its behalf. Project-specific processors, locations and controls are addressed in the applicable project or service agreement. ## Retention and security Data is kept only for as long as reasonably needed for the purpose for which it was collected, ongoing correspondence or work, security and incident handling, applicable limitation periods, and statutory administration. Different records therefore have different retention periods. DDQ uses access controls, data minimisation, maintained systems and other proportionate organisational and technical measures. No internet service can promise absolute security; suspected vulnerabilities can be reported using the contact in [security.txt](https://ddq.nl/.well-known/security.txt). ## Your rights Under applicable data-protection law, you may have rights to access, correct or erase personal data, restrict or object to processing, and receive data in a portable form. Where processing depends on consent, you may withdraw that consent without affecting earlier processing. Rights can depend on the circumstances and may be limited by legal obligations or the rights of others. Send a request to [web@ddq.nl](mailto:web@ddq.nl); DDQ may ask for information needed to verify identity before disclosing or changing data. If a concern cannot be resolved directly, you may lodge a complaint with the Dutch supervisory authority, the [Autoriteit Persoonsgegevens](https://autoriteitpersoonsgegevens.nl/), or another competent data-protection authority. This notice may be updated when the site or applicable practices change; the date above identifies the current version. ## Other sites and customer services This site links to independent websites, research projects and DDQ-operated services. Their own notices apply when you follow those links or sign in to a service. For commissioned systems, DDQ may act as a processor while the customer remains controller; those roles, instructions, retention rules and safeguards are defined for that engagement rather than by this public-site notice. --- Source: https://ddq.nl/privacy/ Site index for agents: https://ddq.nl/llms.txt --- DDQ-project · JTF # Kunststof kansstromen Matching van kunststofreststroom aan toepassing: van kunststofreststroom naar hoogwaardige circulaire producten. [[image: Medegefinancierd door de Europese Unie]](https://ec.europa.eu/regional_policy/information-sources/logo-download-center_en) [[image: Projectposter Kunststof kansstromen met de vijf projectstappen, beoogde impact, samenwerkingspartners en financiers]](https://ddq.nl/assets/20260805%20Poster%20Kunststof%20kansstromen%20%28V1.0%29.png) Klik op de poster om de volledige versie te openen. Het project ## Circulair, meetbaar en digitaal Het consortium PP Recycling, DDQ en Adilanti ontwikkelt een innovatief recycleproces en digitale infrastructuur voor het omzetten van gemengde kunststofafvalstromen in hoogwaardige, rendabele producten. Door een combinatie van (digitale) technieken worden mengstromen efficiënt verwerkt, wat leidt tot lagere kosten, transparantie en economische waarde. Het project stimuleert circulaire ketens, draagt bij aan minder virgin kunststofgebruik en realiseert een significante CO₂-reductie. Met multidisciplinaire expertise, samenwerkingen en opschaling in Zuid-Limburg versterkt het project de regionale economie en versnelt het de transitie naar een duurzame samenleving. - Innovatie - Circulariteit - Samenwerking - Toekomstbestendig Onze aanpak ## Van reststroom naar circulaire plank in vijf stappen 1 ### Kunststof reststromen 2 ### Slimme identificatie Spectropolarimetrie en AI 3 ### Digitaal productpaspoort 4 ### Innovatieve extrusie 5 ### Circulaire planken Duurzame impact ## Minder CO₂, minder afval, meer waarde **25.000 ton** CO₂-reductie **25–30** Nieuwe arbeidsplaatsen **Circulair** Bijdrage aan een circulaire economie Samenwerking ## Projectpartners - PP Recycling - DDQ Pocket Science - Adilanti Skills Innovation Mogelijk gemaakt door ## JTF - Fonds voor een rechtvaardige transitie - Provincie Limburg - Medegefinancierd door de Europese Unie --- Source: https://ddq.nl/jtf/ Site index for agents: https://ddq.nl/llms.txt --- DDQ Cloud # Infrastructure and AI under your control. DDQ Cloud is our privately operated environment for identity, collaboration, code, data, applications and AI. It runs on infrastructure DDQ owns and controls in the Netherlands. Existing users: [**sign in through DDQ Cloud**](https://auth.ddq.nl/). Why it exists ## Autonomy is an operational property. Important knowledge and workflows should not become unusable because an external provider changes its price, product, model or access policy. DDQ Cloud lets us operate critical services and private AI with direct control over the hardware, software and data flows. It is not a claim of complete technological independence. It is a deliberate reduction of avoidable dependency, backed by people who can inspect and operate the complete environment. Private AI ## Models close to the knowledge they use. Two dedicated AI machines serve private inference and persistent agents. DeepSeek Flash provides local model inference. Hermes Agent is the open-source, provider-independent agent framework that supplies memory, tools and autonomous task execution; it is not a model. For customers, we select the model, agent framework, tools and deployment separately according to the task, data sensitivity, performance and governance requirements. **Customer deployment:** private AI can run on customer hardware, inside a customer-controlled environment or on infrastructure operated by DDQ. Permissions, logging, data retention and human control are defined for the actual use case. Live services ## One identity, practical tools. Availability is published in [machine-readable JSON](https://ddq.nl/status.json) and checked every five minutes. - [Identity and single sign-on](https://auth.ddq.nl/)Authelia identity for access to DDQ Cloud services. - [Files, office and calls](https://cloud.ddq.nl/)Nextcloud collaboration and document environment. - [Team communication](https://mattermost.ddq.nl/)Mattermost for project and team conversations. - [Private AI assistant](https://chat.ddq.nl/)Open-source interface to privately operated AI. - [Code repositories](https://git.ddq.nl/)Source control under DDQ administration. - [Knowledge base](https://wiki.ddq.nl/)Structured internal and project knowledge. - [Webmail](https://webmail.ddq.nl/)Email access for DDQ-hosted accounts. Security and control ## Private does not merely mean self-hosted. Strong authentication, controlled access, isolation, encryption, deliberate data flows, recovery and active operation all matter. GDPR/AVG, data minimisation and the EU AI Act are considered as architecture and governance requirements for customer systems. DDQ’s experience operating medical and scientific systems informs this work. Controls are selected for the real risk and responsibility of each service rather than copied from a generic checklist. For organisations ## Use the complete platform or only the capability you need. DDQ can deploy private AI, agents, collaboration services or an application-specific environment on your infrastructure or ours. The first conversation starts with the intended work, users, data sensitivity and required autonomy, not a predetermined product package. For the founder’s practical background and fallback setup, see [noodpakket.tech](https://noodpakket.tech/). To discuss an organisational deployment, contact [web@ddq.nl](mailto:web@ddq.nl). --- Source: https://ddq.nl/cloud.html Site index for agents: https://ddq.nl/llms.txt --- AI process automation # Hand the repetitive work to an agent. A lot of skilled time disappears into copying data between systems, reading and sorting documents, filling in portals and chasing the same information again and again. We build autonomous agents that do that work, so your people can spend their time where their judgement matters. For the right process, that saves thousands of hours of labour. ## What an agent can do - Read and interpret emails, documents and forms. - Use browsers and existing software, also where no API exists. - Call APIs and update your systems. - Complete bounded multi-step workflows, and hand over to a person when a decision is needed. ## How we make it stick Software alone rarely fixes a process. We first learn how the work really happens and who is involved. Then we build the automation, validate it with you, and stay involved through operation and maintenance. Permissions, logging and human control are agreed for each system before anything goes live. ## Any model, no lock-in We use whatever fits the task: a local model on your own hardware when data is sensitive, an external model from a provider such as OpenAI or Anthropic when that works better, plain deterministic software where AI isn't needed, or a controlled mix. Our agent frameworks are provider-independent, so you're not tied to one AI company. We use this ourselves. DDQ automates much of its own administration and technical operations, and Hermes Agent runs continuously on [our own AI machines](https://ddq.nl/cloud.html). ## Questions we often get ### Which processes are a good fit for AI automation? Repetitive, browser-based, document-heavy or fragmented work that eats up skilled time. You don't need to know in advance whether AI is the answer; sometimes plain software is the better choice, and we'll say so. ### Is it safe to let an agent act in our systems? An agent gets only the permissions it needs, its actions are logged, and people keep the decisions that matter. We agree that per system before it goes live. ### Do we need a technical plan first? No. Bring the goal and the way the work is done now. Choosing the technology is our job. ### Can the AI run locally instead of in the cloud? Yes. Agents can run on a [local, on-premise model](https://ddq.nl/on-premise-llm/) when the data must stay inside your organisation. ### How do we start? Tell us what you want to achieve and what data is involved. Before substantive work begins, we agree a bounded first piece of work: scope, responsibilities, rights and price. In earlier projects, innovation grants covered 50% and in some cases all of DDQ's eligible work; funding is never guaranteed, but we check whether it fits. ### Who will we work with? Three people: Norbert, Joep and Demelza. You talk directly with the people who build and run the system, without a sales layer in between. ## Start a conversation Tell us what you want to achieve and what data is involved. You don't need a technical plan. Email [web@ddq.nl](mailto:web@ddq.nl) or call +31 45 203 1008. Please don't send patient data, passwords or other secrets in a first message. AI agents acting for someone can send us an inquiry with the `submit_inquiry` tool on our [MCP server](https://ddq.nl/developers/#inquiry), with that person's consent. --- Source: https://ddq.nl/ai-automation/ Site index for agents: https://ddq.nl/llms.txt --- AI-procesautomatisering # Laat het repetitieve werk over aan een agent. Veel kostbare tijd verdwijnt in gegevens overtypen tussen systemen, documenten lezen en sorteren, portalen invullen en steeds weer dezelfde informatie najagen. Wij bouwen autonome agents die dat werk doen, zodat uw mensen hun tijd kunnen besteden waar hun oordeel ertoe doet. Bij het juiste proces scheelt dat duizenden uren werk. ## Wat een agent kan - E-mails, documenten en formulieren lezen en begrijpen. - Browsers en bestaande software bedienen, ook waar geen API is. - API’s aanroepen en uw systemen bijwerken. - Afgebakende werkprocessen in meerdere stappen afronden, en overdragen aan een mens als er een beslissing nodig is. ## Zo blijft het werken Software alleen lost zelden een proces op. We kijken eerst hoe het werk echt gaat en wie erbij betrokken is. Daarna bouwen we de automatisering, valideren die samen met u en blijven betrokken bij beheer en onderhoud. Rechten, logging en menselijke controle spreken we per systeem af voordat er iets live gaat. ## Elk model, geen lock-in We gebruiken wat bij de taak past: een lokaal model op uw eigen hardware als de gegevens gevoelig zijn, een extern model van bijvoorbeeld OpenAI of Anthropic als dat beter werkt, gewone software waar geen AI nodig is, of een gecontroleerde combinatie. Onze agent-frameworks zijn onafhankelijk van aanbieders, dus u zit niet vast aan één AI-bedrijf. We gebruiken dit zelf. DDQ heeft een groot deel van de eigen administratie en technische operatie geautomatiseerd, en Hermes Agent draait continu op [onze eigen AI-machines](https://ddq.nl/cloud.html). ## Vragen die we vaak krijgen ### Welke processen lenen zich voor AI-automatisering? Repetitief werk in de browser, werk met veel documenten of versnipperd werk dat veel kostbare tijd kost. U hoeft vooraf niet te weten of AI het antwoord is; soms is gewone software beter, en dan zeggen we dat. ### Is het veilig om een agent in onze systemen te laten werken? Een agent krijgt alleen de rechten die hij nodig heeft, zijn acties worden gelogd en mensen houden de beslissingen die ertoe doen. Dat spreken we per systeem af voordat het live gaat. ### Hebben we eerst een technisch plan nodig? Nee. Neem het doel mee en hoe het werk nu gaat. De technologie kiezen is ons werk. ### Kan de AI lokaal draaien in plaats van in de cloud? Ja. Agents kunnen draaien op een [lokaal taalmodel](https://ddq.nl/nl/lokale-llm/) als de gegevens binnen uw organisatie moeten blijven. ### Hoe beginnen we? Vertel ons wat u wilt bereiken en om welke gegevens het gaat. Voordat het echte werk begint, spreken we een afgebakende eerste opdracht af: scope, verantwoordelijkheden, rechten en prijs. In eerdere projecten dekte innovatiesubsidie 50% en soms al het subsidiabele werk van DDQ. Financiering is nooit gegarandeerd, maar we kijken altijd of het past. ### Met wie werken we? Drie mensen: Norbert, Joep en Demelza. U praat rechtstreeks met de mensen die het systeem bouwen en beheren, zonder verkooplaag ertussen. ## Begin een gesprek Vertel ons wat u wilt bereiken en om welke gegevens het gaat. Een technisch plan is niet nodig. Mail naar [web@ddq.nl](mailto:web@ddq.nl) of bel +31 45 203 1008. Stuur in een eerste bericht geen patiëntgegevens, wachtwoorden of andere geheimen. AI-agents die namens iemand handelen, kunnen ons met diens toestemming een aanvraag sturen via de tool `submit_inquiry` van onze [MCP-server](https://ddq.nl/developers/#inquiry). --- Source: https://ddq.nl/nl/ai-automatisering/ Site index for agents: https://ddq.nl/llms.txt --- On-premise LLM # A language model that stays inside your building. Many organisations would love to use AI on their own documents, but can't send patient files, contracts or internal knowledge to a public AI platform. We solve that by running open-weight language models on your own hardware, or on servers DDQ owns and operates in the Netherlands, and connecting them to your own knowledge with retrieval-augmented generation (RAG). ## What you get A private AI assistant that works with your own documents and systems, under your own access rules. The model, the knowledge base and the logs stay on infrastructure you control. When it's justified, we evaluate and fine-tune the model for your field. We deploy and adapt existing open-weight models; we don't train foundation models from scratch. ## Already in use in medicine Our [Medical Copilot™](https://medicalcopilot.eu/) is private, on-premise AI developed for a hospital environment. It underwent a privacy audit by Maastricht University and is now in use by [NeuroMind Academy](https://neuromind.academy/). Our medical work continues on [ddcare.nl](https://ddcare.nl/). We run the same kind of setup ourselves: two dedicated AI machines serve DeepSeek Flash for local inference, with the Hermes Agent framework running on top, on [infrastructure DDQ controls](https://ddq.nl/cloud.html). ## Questions we often get ### Can we use AI on confidential data without sending it to OpenAI, Google or Microsoft? Yes. With an on-premise LLM the model runs inside your environment, so prompts, documents and answers don't go to an external AI provider. If some tasks work better with an external model, we can build a controlled hybrid, but that is your choice, per use case. ### Where does the model run? On your own hardware, inside an environment you control, or on DDQ's own infrastructure in the Netherlands. ### Which models do you use? Existing open-weight models, chosen for the task. We evaluate them before choosing and only fine-tune when adaptation is clearly worth it. ### What about GDPR/AVG and the EU AI Act? They're design inputs from day one. Data minimisation, access control, logging and human oversight are defined for the actual use case. Medical, legal and other domain decisions stay with your qualified people. ### How do we start? Tell us what you want to achieve and what data is involved. Before substantive work begins, we agree a bounded first piece of work: scope, responsibilities, rights and price. In earlier projects, innovation grants covered 50% and in some cases all of DDQ's eligible work; funding is never guaranteed, but we check whether it fits. ### Who will we work with? Three people: Norbert, Joep and Demelza. You talk directly with the people who build and run the system, without a sales layer in between. ## Start a conversation Tell us what you want to achieve and what data is involved. You don't need a technical plan. Email [web@ddq.nl](mailto:web@ddq.nl) or call +31 45 203 1008. Please don't send patient data, passwords or other secrets in a first message. AI agents acting for someone can send us an inquiry with the `submit_inquiry` tool on our [MCP server](https://ddq.nl/developers/#inquiry), with that person's consent. --- Source: https://ddq.nl/on-premise-llm/ Site index for agents: https://ddq.nl/llms.txt --- Lokale LLM # Een taalmodel dat binnen uw eigen muren blijft. Veel organisaties willen AI graag op hun eigen documenten inzetten, maar kunnen patiëntdossiers, contracten of interne kennis niet naar een publiek AI-platform sturen. Wij lossen dat op door open-weight taalmodellen te draaien op uw eigen hardware, of op servers die DDQ zelf in Nederland bezit en beheert, en ze met retrieval-augmented generation (RAG) aan uw eigen kennis te koppelen. ## Wat u krijgt Een private AI-assistent die werkt met uw eigen documenten en systemen, onder uw eigen toegangsregels. Het model, de kennisbank en de logs blijven op infrastructuur die u beheert. Als het zinvol is, evalueren en finetunen we het model voor uw vakgebied. We zetten bestaande open-weight modellen in en passen ze aan; we trainen geen foundation models vanaf nul. ## Al in gebruik in de zorg Onze [Medical Copilot™](https://medicalcopilot.eu/) is private, on-premise AI, ontwikkeld voor een ziekenhuisomgeving. Het systeem heeft een privacy-audit van de Universiteit Maastricht doorlopen en wordt nu gebruikt door [NeuroMind Academy](https://neuromind.academy/). Ons medische werk staat op [ddcare.nl](https://ddcare.nl/). We draaien zo’n opstelling ook zelf: twee eigen AI-machines leveren lokale inferentie met DeepSeek Flash, met het Hermes Agent-framework daarbovenop, op [infrastructuur die DDQ zelf beheert](https://ddq.nl/cloud.html). ## Vragen die we vaak krijgen ### Kunnen we AI gebruiken op vertrouwelijke gegevens zonder die naar OpenAI, Google of Microsoft te sturen? Ja. Bij een lokale LLM draait het model binnen uw eigen omgeving, dus vragen, documenten en antwoorden gaan niet naar een externe AI-aanbieder. Werken sommige taken beter met een extern model, dan kunnen we een gecontroleerde hybride bouwen, maar dat is per toepassing uw keuze. ### Waar draait het model? Op uw eigen hardware, in een omgeving die u zelf beheert, of op de eigen infrastructuur van DDQ in Nederland. ### Welke modellen gebruiken jullie? Bestaande open-weight modellen, gekozen voor de taak. We evalueren ze vooraf en finetunen alleen als aanpassing echt de moeite waard is. ### Hoe zit het met de AVG en de EU AI Act? Die zijn vanaf dag één ontwerpeisen. Dataminimalisatie, toegangsbeheer, logging en menselijk toezicht leggen we vast voor de werkelijke toepassing. Medische, juridische en andere inhoudelijke beslissingen blijven bij uw eigen deskundigen. ### Hoe beginnen we? Vertel ons wat u wilt bereiken en om welke gegevens het gaat. Voordat het echte werk begint, spreken we een afgebakende eerste opdracht af: scope, verantwoordelijkheden, rechten en prijs. In eerdere projecten dekte innovatiesubsidie 50% en soms al het subsidiabele werk van DDQ. Financiering is nooit gegarandeerd, maar we kijken altijd of het past. ### Met wie werken we? Drie mensen: Norbert, Joep en Demelza. U praat rechtstreeks met de mensen die het systeem bouwen en beheren, zonder verkooplaag ertussen. ## Begin een gesprek Vertel ons wat u wilt bereiken en om welke gegevens het gaat. Een technisch plan is niet nodig. Mail naar [web@ddq.nl](mailto:web@ddq.nl) of bel +31 45 203 1008. Stuur in een eerste bericht geen patiëntgegevens, wachtwoorden of andere geheimen. AI-agents die namens iemand handelen, kunnen ons met diens toestemming een aanvraag sturen via de tool `submit_inquiry` van onze [MCP-server](https://ddq.nl/developers/#inquiry). --- Source: https://ddq.nl/nl/lokale-llm/ Site index for agents: https://ddq.nl/llms.txt --- Autonomous cloud # Your own cloud, whatever Big Tech decides next. Most organisations run on a handful of American cloud services. That works fine, until a price, a policy, an account or a political decision changes. That's no reason to panic, but it is common sense to have your own. We build autonomous clouds on open-source software, on your hardware or on servers DDQ owns and operates in the Netherlands. ## What it can include - Files, office documents and video calls (Nextcloud). - Team chat (Mattermost). - A private AI assistant, running on local models. - Email and webmail. - Git code repositories. - A wiki for shared knowledge. - Single sign-on for all of it (Authelia). ## We run it ourselves This is how DDQ works every day. [DDQ Cloud](https://ddq.nl/cloud.html) runs on multiple servers and two dedicated AI machines that we own and control in the Netherlands, and its [service status](https://ddq.nl/status.json) is checked every five minutes. It isn't a claim of complete technological independence. It's a deliberate reduction of avoidable dependency, run by people who can inspect and operate the whole environment. Background and practical setup notes: [noodpakket.tech](https://noodpakket.tech/). ## Questions we often get ### Can this replace Microsoft 365 or Google Workspace? It covers files, office documents, chat, email, code, a wiki and single sign-on with open-source software. Which services you move, and when, we decide together; you don't have to switch everything at once. ### Where is our data stored? On your own hardware, or on DDQ-owned servers in the Netherlands. Either way the data stays in the EU. ### Can we run private AI in it? Yes. Local language models can run right next to your data. See [on-premise LLM](https://ddq.nl/on-premise-llm/). ### Is open source secure enough? Security depends on how a system is built and run. We use strong authentication with single sign-on, keep the software maintained, and monitor the services continuously. ### How do we start? Tell us what you want to achieve and what data is involved. Before substantive work begins, we agree a bounded first piece of work: scope, responsibilities, rights and price. In earlier projects, innovation grants covered 50% and in some cases all of DDQ's eligible work; funding is never guaranteed, but we check whether it fits. ### Who will we work with? Three people: Norbert, Joep and Demelza. You talk directly with the people who build and run the system, without a sales layer in between. ## Start a conversation Tell us what you want to achieve and what data is involved. You don't need a technical plan. Email [web@ddq.nl](mailto:web@ddq.nl) or call +31 45 203 1008. Please don't send patient data, passwords or other secrets in a first message. AI agents acting for someone can send us an inquiry with the `submit_inquiry` tool on our [MCP server](https://ddq.nl/developers/#inquiry), with that person's consent. --- Source: https://ddq.nl/autonomous-cloud/ Site index for agents: https://ddq.nl/llms.txt --- Autonome cloud # Uw eigen cloud, wat Big Tech ook besluit. De meeste organisaties draaien op een handvol Amerikaanse clouddiensten. Dat gaat prima, tot er een prijs, een voorwaarde, een account of een politieke beslissing verandert. Dat is geen reden voor paniek, maar het is wel gewoon verstandig om een eigen alternatief te hebben. Wij bouwen autonome clouds op open-source software, op uw eigen hardware of op servers die DDQ in Nederland bezit en beheert. ## Wat erin kan zitten - Bestanden, office-documenten en videobellen (Nextcloud). - Teamchat (Mattermost). - Een private AI-assistent op lokale modellen. - E-mail en webmail. - Git-repositories voor code. - Een wiki voor gedeelde kennis. - Eén login voor alles (Authelia). ## We gebruiken het zelf Zo werkt DDQ elke dag. [DDQ Cloud](https://ddq.nl/cloud.html) draait op meerdere servers en twee eigen AI-machines die we in Nederland bezitten en beheren, en de [status van de diensten](https://ddq.nl/status.json) wordt elke vijf minuten gecontroleerd. Het is geen claim van volledige technologische onafhankelijkheid. Het is een bewuste vermindering van vermijdbare afhankelijkheid, beheerd door mensen die de hele omgeving kunnen doorgronden en bedienen. Achtergrond en praktische installatie-notities: [noodpakket.tech](https://noodpakket.tech/). ## Vragen die we vaak krijgen ### Kan dit Microsoft 365 of Google Workspace vervangen? Het dekt bestanden, office-documenten, chat, e-mail, code, een wiki en single sign-on met open-source software. Welke diensten u overzet, en wanneer, bepalen we samen; u hoeft niet alles tegelijk over te zetten. ### Waar staan onze gegevens? Op uw eigen hardware, of op servers van DDQ in Nederland. In beide gevallen blijven de gegevens in de EU. ### Kunnen we er private AI in draaien? Ja. Lokale taalmodellen kunnen direct naast uw gegevens draaien. Zie [lokale LLM](https://ddq.nl/nl/lokale-llm/). ### Is open source veilig genoeg? Veiligheid hangt af van hoe een systeem gebouwd en beheerd wordt. We gebruiken sterke authenticatie met single sign-on, houden de software bij en bewaken de diensten continu. ### Hoe beginnen we? Vertel ons wat u wilt bereiken en om welke gegevens het gaat. Voordat het echte werk begint, spreken we een afgebakende eerste opdracht af: scope, verantwoordelijkheden, rechten en prijs. In eerdere projecten dekte innovatiesubsidie 50% en soms al het subsidiabele werk van DDQ. Financiering is nooit gegarandeerd, maar we kijken altijd of het past. ### Met wie werken we? Drie mensen: Norbert, Joep en Demelza. U praat rechtstreeks met de mensen die het systeem bouwen en beheren, zonder verkooplaag ertussen. ## Begin een gesprek Vertel ons wat u wilt bereiken en om welke gegevens het gaat. Een technisch plan is niet nodig. Mail naar [web@ddq.nl](mailto:web@ddq.nl) of bel +31 45 203 1008. Stuur in een eerste bericht geen patiëntgegevens, wachtwoorden of andere geheimen. AI-agents die namens iemand handelen, kunnen ons met diens toestemming een aanvraag sturen via de tool `submit_inquiry` van onze [MCP-server](https://ddq.nl/developers/#inquiry). --- Source: https://ddq.nl/nl/autonome-cloud/ Site index for agents: https://ddq.nl/llms.txt --- Source: https://ddq.nl/notes/2026-06-11-agent-native-ddq-nl # Making ddq.nl agent-native ddq.nl runs on infrastructure operated by DDQ. We rebuilt the site so its public information is consistent for people, crawlers and software agents. ## What changed - **One source of truth.** The site's facts live in a canonical JSON file. A build script generates `llms.txt`, Markdown content, `/api/company.json`, the sitemap, security information and agent-discovery documents. - **MCP endpoint.** AI agents can query the site over the Model Context Protocol at [/mcp/](/mcp/) using Streamable HTTP without authentication. The tools return the company profile, cloud services with live status, and contact information. The endpoint is implemented in plain PHP without application dependencies. - **Live status.** The [cloud page](/cloud.html) now shows the actual up/down state of our self-hosted services, checked every five minutes from this machine. No third-party status service. - **Discovery and security.** The site publishes `AGENTS.md`, OpenAPI, WebMCP and MCP discovery, an API catalog, `security.txt` (RFC 9116), WebFinger, structured data and explicit crawler policy. ## Why DDQ develops on-premise and self-hosted systems. The website uses the same approach: machine-readable interfaces, observable service status and infrastructure operated by DDQ.