When does a custom AI system make sense?
Start with the work you or your organization need to finish. A custom system is worth discussing when that work depends on your own context, knowledge, or connected tools, and a general-purpose AI tool does not fit the task.
One possible project is an Arabic/English assistant that answers from permitted company documents, reads an authorized CRM record and drafts a follow-up task for approval. The knowledge sources, access rules and proposed actions need to be scoped against your actual systems.
A simpler workflow or an existing product may already address the need. Include what you have tried and where it falls short so the conversation starts with the actual gap.
How can Dahshan Labs help with your AI project?
Start with consultation, system design or a custom build. The right scope depends on the problem, available knowledge and tools, and the people who will use the system.
Explore the bilingual assistant architecture and runnable example
- AI consultation and system design
- Select a useful AI use case, assess feasibility against your data and constraints, and outline the system design. Agree a practical next step, whether that is evaluating an existing tool, preparing a pilot or refining the requirements.
- Internal knowledge and project memory
- Keep decisions, corrections and evidence available to the AI tools that need them. KLYPIX demonstrates project-scoped retrieval and shared context across supported development tools.
- Custom AI agents and controlled tool use
- Define the job, required knowledge and allowed tools, then inspect the run. AgentMug calls these portable agents workers; its open core documents worker packages, source-readiness checks, runtime policies and run receipts.
- MCP and workflow integrations
- Make AI useful inside an existing workflow. Our public projects include MCP integrations and an AgentMug n8n node. Connecting a particular CRM, ERP or internal API depends on its interface, access rules and the actions you need.
What does Dahshan Labs build?
KLYPIX and AgentMug are products we build ourselves. The engineering notes below explain the problem, the implementation and its limits, with links to public documentation and source code.
KLYPIX
KLYPIX is a portable project workspace with a shared brain. It keeps files, evidence, decisions, corrections, and current context together for people and supported AI tools.
How we build shared project knowledge with KLYPIX
Explore KLYPIX and its documentation
AgentMug
AgentMug lets you build portable AI workers, connect them to real systems and trusted knowledge, and run them wherever the work happens, with a clear record of each run.
How we build portable AI workers with AgentMug
View the AgentMug open core on GitHub
KLYPIX and AgentMug are available products. A custom project starts with your requirements; the right approach depends on the work, systems, and constraints you describe.
What should a useful first build prove?
Start with one workflow, one representative knowledge source and one connected system. A useful brief names the questions to answer and the action to complete, then makes success observable:
Explore the bilingual assistant architecture and runnable example
- Trace an answer to the source or project decision behind it, and make missing or stale information visible.
- Exercise the access rules with different users and show what happens when a requested action is not allowed.
- Separate reading information from changing a business record. Identify the changes that need a person's approval.
- Review actual runs against agreed examples, including failures and Arabic or English inputs where required.
- Agree where the data and models run, what gets recorded, and what you or your team need to operate the system.
What should your project brief include?
A short, concrete brief is enough to start. Useful details include:
- The work
- Who needs the system, what they do today, and where the work becomes unclear or repetitive.
- The systems and knowledge
- The tools or knowledge sources involved, and which information the AI would need to use.
- The boundaries
- Language needs, access limits, hosting requirements, and steps that should remain under human review.
- The useful result
- What a successful first step would look like, with your target timing and budget if known.
Use a high-level description or a redacted example. Please do not include passwords, access keys, or sensitive data in the contact form.
How we handle contact dataCommon questions
Where is Dahshan Labs based?
Dahshan Labs is a frontier AI lab based in Riyadh, Saudi Arabia.
Who are your AI services for?
We offer AI system design, development and consultation for individuals, companies and enterprises. A personal workflow, a company tool and an enterprise integration have different requirements; we scope the work around your users, systems and constraints.
Can I request a standalone AI consultation?
Yes. You can discuss a consultation without committing to a development project. We can help with use-case selection, feasibility, system design and a practical next step. Share the decision you need to make and the constraints you already know.
Can you scope an Arabic/English assistant for our ERP or CRM?
Yes. Start with one knowledge source, one system and the actions your users need. Our integration guide explains a proposed architecture and provides a synthetic access-and-approval example; connecting your environment and validating bilingual answer quality are separate pilot steps.
Can I discuss a project in Arabic or English?
Yes. The website and project contact form are available in Arabic and English. Use the language that best explains your project.
How much does custom AI development cost, and how long does it take?
This site does not publish fixed project prices or delivery timelines. Share the scope, connected systems, constraints, and intended result so the next step can be discussed for your project.
Can we connect AI to our documents and business tools?
Specify the knowledge sources, systems involved, and whether the AI should only read or also perform actions. Our products demonstrate project knowledge, MCP integrations and portable AI workers; your systems still need an integration assessment.
Does every knowledge project need RAG or a vector database?
No. The retrieval method should match the sources and questions. KLYPIX documents lexical retrieval and optional semantic retrieval. AgentMug defines source and knowledge-adapter interfaces; its open core does not ship a universal vector-search implementation.
Can we require private hosting and source-code ownership?
Include hosting, data access, source-code ownership and handover requirements in the brief so they can be agreed for your project. The licenses of our public product components are separate from the terms of a custom engagement.
How do I contact Dahshan Labs?
Send a project brief through the contact form. An email address is required for a reply; a phone number is optional.