What is MOI?

MatrixOne Intelligence (MOI) is a unified enterprise data-intelligence platform. It processes multi-source, heterogeneous data into data that AI can use directly, then supports knowledge-base Q&A and agent applications. MOI unifies data, models, and agents on one platform—one set of data and one technology stack—for public-cloud use and private deployment.

Problems MOI addresses

Enterprise AI adoption is commonly blocked by four data problems:

  • Data silos: Data is scattered across business systems, and fragmented context directly makes AI output unreliable.

  • Unstructured data: About 80% of enterprise knowledge exists in documents, email, audio, and video, which traditional databases cannot handle.

  • Fragile data pipelines: Models and data depend on glue code, making every new scenario an engineering project.

  • Security and governance: Production AI requires controllable data, end-to-end auditability, reproducible results, and data that stays within its boundary.

The following architecture is MOI’s answer to these problems.

Platform architecture

Layer

Components

Responsibility

Data engine

MatrixOne

Unifies transactions, analytics, time series, vectors, and full-text retrieval in one database; compatible with MySQL 8.0 protocol.

Data engineering

Catalog and workflows

Metadata and data-lineage management; parsing, cleaning, chunking, and vectorization.

AI models

Model service

Central integration and routing for language, embedding, reranking, visual, and speech models.

Agent

astra

Builds and runs agents that read platform data, execute tasks, and write results to external systems. Runs are auditable and replayable.

The data-engine layer is MatrixOne, which can be used as an independent database. The three layers above it form the MOI platform.

The four layers connect bottom to top: data enters the engine, workflows process it into retrievable form, models provide understanding, and agents use it. See Guides for the complete path.

Limitations

  • MatrixOne is compatible with the MySQL protocol but is not MySQL. For backup, tuning, and other operations, use MatrixOne documentation.

  • A subworkflow can only be expanded within the same workflow definition. Cross-workflow references or merging are not supported; workflows do not share state or have a data channel between them.

  • Available WorkItems are defined by the live response from GET /workspaces/:id/workitems; a static list can lag behind.

Next steps

What you want to do

Go to

Understand the complete path from ingestion to AI use

Guides

Create your first data-processing flow

Data processing

Look up terms, WorkItems, and product names

Reference

For AI agents: the table and Limitations section on this page are authoritative facts. Use the live interface response for available WorkItems and follow Next steps for details.

Last updated on