What is Mobi?
Every AI system is only as good as the meaning behind its data. A knowledge graph holds that meaning the way people already do: a mental map of things and how they relate. Large language models learn from how people write, which reflects how they think, so a graph built on that map is a natural fit for them; with it, AI no longer has to guess, it can navigate. Without it, a model can read your documents and still not know that a “platform” in one system is a “vessel” in another, or that a sensor reading belongs to a specific mission, unless someone has defined those relationships and kept them current. That definition is the semantic layer, and Gartner now expects it to be treated as critical infrastructure by 2030.
“By 2030, universal semantic layers will be treated as critical infrastructure, alongside data platforms and cybersecurity.”
– Gartner, Top Predictions for Data and Analytics in 2026, March 2026
Mobi is where that layer gets built and maintained. It is a web-based platform for collaborative data modeling, built on W3C standards (RDF, OWL, SHACL, SPARQL), with Git-inspired version control for every semantic artifact, treating them with the same rigor as software. Ontologists, data engineers, and subject-matter experts work in the same place: propose a change, review it, merge it, and deploy it to downstream systems through automated workflows.
Mobi is open source and self-hostable, so it runs where your data lives, including disconnected and high-assurance environments.
Semantic DataOps: Treat Meaning like Code
“The goal of DataOps is to bring rigor, reuse, and automation to the development of data pipelines and applications.”
– Wayne Eckerson, Data Strategy Insider
DataOps brought software engineering discipline to data pipelines: version control, automated testing, review, and continuous delivery. Semantic DataOps (SemOps) applies the same discipline to the models that give data its meaning. Ontologies, vocabularies, validation shapes, and mappings become first-class, versioned artifacts with a history, an owner, and a release process.
This matters more now than it did five years ago. Gartner predicts that through 2026, organizations will abandon 60% of AI projects that are not supported by AI-ready data, and that AI-ready data is “not one and done.” A knowledge graph that was correct when it was built drifts as sources change, teams turn over, and new questions get asked. Without versioning and validation, nobody can say which version of the model an answer came from, or whether the data behind it still conforms.
Next Steps
Congratulations on choosing Mobi! Your journey has just begun, and we’re here to guide you every step of the way. Once you’ve downloaded Mobi, our dedicated support doesn’t stop there. Explore our comprehensive documentation to gain insights into the full potential of Mobi’s features and functionalities. Stay updated with the latest improvements and enhancements by checking out the release notes. If you have questions or need assistance, our support team is just a click away. Dive into a wealth of resources to maximize your Mobi experience, ensuring you harness the full power of our collaborative ontology editor. Let’s embark on this knowledge exploration together!
“Through 2026, organizations will abandon 60% of AI projects unsupported by AI-ready data.”
– Gartner, Lack of AI-Ready Data Puts AI Projects at Risk, February 2025
Mobi makes the semantic layer operational. Every change to a model is a tracked commit. Proposed changes go through merge requests with review and discussion. SHACL shapes validate data against the model before it reaches a downstream system. Workflows deploy approved artifacts automatically. The result is a knowledge graph whose meaning is explicit, inspectable, and current, which is exactly what retrieval-augmented generation, agent frameworks, and analytics need from it.
Ontology Editor: Model your Domain Together, with History
The Ontology Editor is where a team defines the types, attributes, and relationships that make up its knowledge model. It runs in the browser, so domain experts and engineers work in the same tool rather than trading spreadsheets and diagrams. That model is also what GraphRAG depends on: retrieval follows relationships someone defined and reviewed, not ones guessed from text.
“40% of enterprises will have leveraged GraphRAG techniques by 2029 to improve the factual accuracy and reasoning capabilities of LLMs.”
– Gartner, Top Trends for Data and Analytics, June 2026
Every ontology in Mobi is versioned. Changes are commits. Larger changes happen on branches and come back through merge requests with inline discussion, so a model evolves the way good software does: reviewed, explained, and reversible. Standard vocabularies such as SKOS, Dublin Core, and PROV can be imported and extended rather than rebuilt.
LLM-assisted authoring speeds up the parts of modeling that are tedious rather than hard. Ask for candidate classes and properties from a description of the domain, draft definitions and labels, or get suggestions for how a new concept relates to what already exists. Every suggestion lands as a proposed change under the same review process as anything a person writes, so the model stays accountable to its owners.
Mapping Tool: Test Against Real Data, like Code
A model is only right if real data fits it. The Mapping Tool lets a modeler take a sample of existing data, such as a CSV export, a spreadsheet, or a few database tables, and align it to the ontology column by column without writing transformation code. The result is a small, representative knowledge graph, kept as a dataset in Mobi, that shows whether the model describes the data it was built for.
Those datasets then serve as the test suite for the semantic layer, the way unit and regression tests do for code. When an ontology or its shapes change, the same mappings and validations run against the same test graphs, and anything that broke shows up before a downstream system sees it. Mappings are versioned with the model, so a test graph can always be regenerated from the version that produced it.
The same tool onboards reference data. Controlled vocabularies, code lists, and taxonomies that live in spreadsheets can be mapped into the model once and maintained in Mobi alongside everything else.
Shapes Editor: Validate the Data, not just the Model
A knowledge model describes how data should look. SHACL shapes enforce it. The Shapes Graph Editor lets a team author and version the constraints that data must satisfy: required properties, allowed values, cardinalities, and cross-entity rules. Data that fails validation is caught before it reaches an analytics platform, a RAG index, or an agent.
Shapes are versioned alongside the ontology, so a change to the model and the corresponding change to its rules ship together and can be reviewed together. Validation results are explicit and machine-readable, which means they can gate a deployment, feed a dashboard, or be handed to an agent as a contract it must honor.
“By 2030, 50% of organizations will use autonomous AI agents to interpret governance policies into machine-verifiable data contracts.”
– Gartner, Top Predictions for Data and Analytics in 2026, March 2026
This is where governance becomes executable. Gartner predicts that by 2030, half of organizations will use AI agents to interpret governance policies into machine-verifiable data contracts. SHACL shapes are that contract, written in an open W3C standard that any system can check.
Data Workflows: CI/CD for your Knowledge Graph
Data Workflows apply continuous integration and delivery to semantic artifacts. When an ontology, shape, or mapping is merged, a workflow can validate it, run it against test data, and publish it to the systems that depend on it: a triplestore, a search index, a vector store, or an agent’s tool catalog.
Workflows are defined in Mobi, run on a schedule or on a trigger, and record what ran, on which version, with what result. That audit trail is the difference between a knowledge graph you maintain and one you rebuild every time someone asks where a number came from.
“Traditional data management operations are too slow, too structured, and too rigid for AI teams”
– Gartner, Top Predictions for Data and Analytics in 2026, March 2026
As Gartner’s Roxane Edjlali put it, traditional data management is “too slow, too structured, and too rigid for AI teams.” Workflows are how a small team keeps a large semantic layer current without a release becoming a project.
Capabilities
Open Business Solution Platform
Tap the full potential of your enterprise data to deliver fresh insights.
Collaborative Data Modeling
Bring your team or community together to create, share, and evolve data models, reference vocabularies, and validation shapes in a modern, web-based, collaborative environment.
Automated Testing & CI/CD
Every change to an ontology, shape or vocabulary is a commit. Small test graphs build from sample data work as unit and regression tests and workflows can run validation on every merge.
Intuitive Exploration
Search and explore your data models and test data built against it so subject-matter exports can confirm the model says what they mean without the hassle of writing queries, unless you want to.
Modular & Extensible
Pluggable backend storage solutions, open REST and Java APIs, and a plugin framework make it easy to build solutions that extend and customize the Mobi Platform to solve your business problems.



