The questions buyers, architects and admins actually ask.

x101 is an enterprise AI platform built by VisionWaves: a single intelligence layer that sits between your people and all of your existing systems, including ERP, CRM, ITSM, document repositories, databases and custom APIs. It does not replace those systems; it sits across them. Teams ask questions, retrieve knowledge and execute multi-system actions in plain language or voice, and x101 returns sourced answers or completed actions, governed on each request. See the Platform page for the full architecture.

x101 brings agentic orchestration, agentic workflow automation, enterprise knowledge retrieval, deep research, voice agents and a no-code Agent Studio into one platform, all reachable from a single Command Centre. Teams use it across IT, security, CRM, finance, HR and telecom for tasks that span several systems at once. Representative solutions are shown across the case studies.

Each request flows through six governed stages: Ask, Analyse, Plan, Execute, Verify and Deliver. Intent and identity are resolved first, permission checks run before any data is retrieved, work is executed across systems in parallel, the output is verified against source data and policy before it surfaces, and the full chain is captured in an immutable audit trail. Each stage is explained on the Platform page.

x101 is system-agnostic and model-agnostic. Rather than being tied to one vendor's cloud or one large language model, it unifies all enterprise systems under a single governed layer, with RBAC/ABAC enforced before retrieval, PII masking, output guardrails and a signed, immutable audit trail. You choose the deployment environment and the underlying model, so there is no ecosystem or inference lock-in. The comparison on the home page sets it against ecosystem copilots, enterprise search and workflow platforms.

x101 is the enterprise intelligence layer within the VisionWaves family. Several siblings build on or alongside it: OpsSingularity (Sentinel AI) for IT and security operations is built on x101, while Enterprise Singularity provides the broader operating layer, NetSingularity owns telecom OSS/BSS, Code Singularity delivers complete systems, and DataByte handles data engineering. The shared traits across the family are consistent: governed, any cloud or on-premises, no lock-in, with AI agents at the centre.

x101 is not locked to any single LLM provider. It works with Anthropic Claude, OpenAI GPT, Google Gemini, Mistral, Llama and open-weight models running on your own hardware, and the model is swapped through configuration rather than re-architecture. This is possible because the model is a replaceable component inside the intelligence layer, not the platform itself: the four stages that do the work, understand, recall, reason and act, along with orchestration, memory, explainability and cost tracking, all sit around the model rather than inside it. Changing the model therefore does not change how agents are built, how permissions are enforced before retrieval, or how answers are cited and audited. The practical effect is that x101 can adopt newer or more cost-effective models as they appear, and your deployment is not exposed to any single vendor's pricing, availability or roadmap.

Yes. Because x101 is model-agnostic, it can run against on-premises and open-weight models as well as hosted commercial ones. This is what makes a fully air-gapped deployment possible: the inference can stay entirely inside your environment.

The model is a configuration choice, not a hard dependency. Orchestration, retrieval, governance and connectors all sit above the model layer, so swapping or mixing models does not require rebuilding agents or workflows. The investment stays in the platform, not in any one vendor's model.

Answers are grounded in retrieved source content rather than the model's memory alone, and they are returned with citations to the exact document, page and version. Retrieval is built to reduce the chance of a thin answer in the first place: approved content is written to three stores, a vector index of embeddings, a BM25 keyword index and a knowledge graph of entities and relationships, and a question is answered by searching all three in parallel, then fusing and reranking the results into one ranked set. Retrieval runs as a loop rather than a single shot, so the agent judges what came back and searches again when the evidence is thin. Before anything is surfaced, the verify stage cross-checks the output against source data and policy, so unsupported claims are caught before they reach the user, and the citation travelling with each claim lets a reader check the source directly.

Latency and throughput are determined by the model and inference provider in use, the level of reasoning required, and the complexity of the task. As a general guide, simpler queries typically return in around two seconds, while complex operations that need multiple agents to orchestrate and correlate across systems can take up to a few minutes.

x101 uses agentic retrieval-augmented generation (RAG) across repositories such as SharePoint, Confluence, PDFs and wikis. Retrieval is permission-checked against the requesting user, so people only see what they are entitled to see, and each answer is grounded in its source with page-and-version citations that can be opened and verified.

An orchestrator decomposes a task, dispatches a team of specialist agents to work in parallel across different systems, and then synthesises their results into one answer. Coordinating work this way keeps quality high on large, multi-step jobs that would otherwise exceed a single model's context window.

Yes. The no-code Agent Studio is a visual builder where a business user configures an agent's persona, tools, memory and guardrails, and connects it over native MCP and A2A standards. Agents can be built and shipped without writing code, while developers can extend them through the SDK.

Deep Research runs automated, multi-step research that scans many sources in parallel and returns a cited, formatted deliverable in minutes rather than a list of links. Each line of enquiry runs with its own context, which keeps quality uniform across a large body of material instead of degrading as the work grows.

Re-indexing runs on a schedule that you define. Each connected source can be set to refresh on the cadence that suits how often its content changes, so the knowledge base stays as current as your data requires.

Governance is enforced on each request, not bolted on afterwards. It runs across all four stages of the intelligence layer rather than being one of them: identity is resolved before the model sees the text, RBAC/ABAC entitlement checks run before any data is retrieved, PII can be masked in line with policy, and input and output guardrails validate safety on the way in and on the way out. Actions that need a person can be held behind a human approval gate. Each step is recorded in a signed, immutable audit log, with traceability per user, per agent and per action, so a reviewer can replay exactly what happened rather than infer it from search or conversation logs. VisionWaves maintains an information security management system aligned to ISO/IEC 27001:2022, and the Data Processing Agreement sets out the processing terms.

x101 evaluates role-based and attribute-based access control before retrieval, during the Analyse stage, and it acts only within the permissions of the requesting user. A person can never retrieve or act on data they are not entitled to, even through a conversational or voice interface.

The audit trail captures the key metrics for each step of a request, including the tokens used, the cost, the model invoked and the time taken, alongside user-level attributes such as who made the request. Coverage is per user, per agent and per action rather than a general activity or conversation log, so a single request can be reconstructed stage by stage: what was understood, what was retrieved and under whose entitlement, which tools were called, what the verify stage checked, and what was returned. Each action is recorded in a signed, immutable log, built for compliance review and for replaying exactly what happened. Because cost and token usage sit on the same record, one trail answers both the governance question of who did what under which permission, and the finance question of what each user and agent is consuming.

Yes. Conversation data is isolated between tenants and stored in per-client repositories and persistence, so one customer's data is never co-mingled with another's. No customer or user data is used to train the underlying models, and the Data Processing Agreement sets out the processing terms. Where isolation needs to be stronger than an architectural and contractual guarantee, the platform can run entirely inside your own boundary: in a private cloud, on-premises, or fully air-gapped, with the models, the inference and the knowledge they retrieve all running on hardware you own. In that configuration no prompt and no retrieved document crosses the perimeter at all. That is the posture organisations take when data leaving the boundary is itself a compliance violation, including sovereign and government deployments, banks under residency controls, and health systems that treat operational data as patient-data adjacent.

x101 runs on any public cloud (AWS, Azure, GCP), in a private cloud, on-premises, or in a fully air-gapped environment. This makes it suitable for organisations with strict data residency, isolation or sovereignty requirements. Air-gapped here means the whole platform, AI included: orchestration, retrieval, analytics, research, agents and actions all operate indefinitely with zero outbound dependency, and the serving stack runs on GPUs you own in a facility you control, so no prompt crosses the boundary, not for an answer and not for an analysis. It has been shipped that way in production, not only supported on paper. The deployment model does not change what the platform does: the same agents, connectors, governance and audit trail apply whichever way it runs, so deciding where the data lives is not also a decision about which capabilities you give up.

x101 is designed to scale horizontally, and its orchestration model runs specialist agents in parallel. Large, multi-step jobs keep a uniform quality as volume increases, rather than degrading on a single long context.

Because x101 deploys against existing systems, data and governance rather than requiring a rebuild, a pilot can typically start in weeks and then roll out at the organisation's own pace. The exact timeline depends on the number of systems connected and the governance review involved.

x101 runs on any public cloud as well as private environments, deployed on Kubernetes (K8s). This keeps it portable across public cloud, private cloud and on-premises infrastructure without re-architecting.

x101 connects to enterprise applications such as ERP, CRM and ITSM, to SQL and other databases, to document repositories, to email and to SaaS tools, over more than 2,000 connectors and native MCP and A2A standards, plus APIs and SQL. It can read, write and act across these systems within the permissions of the requesting user. Browse the full list on the Connectors page.

MCP (Model Context Protocol) is an open standard for connecting AI models to tools and data sources. A2A (Agent-to-Agent) is an open standard that lets agents discover, talk to and delegate work to each other across vendors. x101 supports both natively, which keeps integrations standards-based and portable rather than proprietary. Both terms are defined in the Glossary.

Yes. Beyond the pre-built connector library, x101 provides an SDK together with native MCP and A2A support for connecting custom, in-house or legacy systems. Coverage is not limited to the systems that ship with the platform.

More than 2,000 connectors are available, spanning enterprise applications, databases, document repositories, communication tools and SaaS services. The current catalogue, grouped by domain and searchable, is on the Connectors page.

x101 ships with a set of platform components, one of which is User Management. It provides role-based access control (RBAC), attribute-based access control (ABAC) and AoA, so administrators can manage users, roles and access policies at scale from within the platform.

Reach out through the Contact page, where the Singapore, USA and India offices are listed by phone and email, and the team typically responds within one business day. Because x101 deploys against existing systems, data and governance, a pilot can usually start in weeks and then roll out at your own pace.