MatrixOne¶

Building on the preceding platform: the engine beneath AI Studio¶

The tables and volumes you create in Catalog, the queries run by the SQL editor and query compute resources, and knowledge base vector indexes in AI Studio all run on a database engine underneath—that engine is MatrixOne. This page describes the foundation itself: what it is, where its capabilities come from, and what you see when you need to use it directly.

Positioning¶

MatrixOne is a hyper-converged, cloud-native database for AI applications. A single engine supports mixed workloads including transactions (OLTP), analytics (OLAP), vector search, and full-text search. Enterprises no longer need to maintain separate systems for an operational database, data warehouse, vector database, and search engine, or repeatedly move data among them.

For most users who work only with data and agents in AI Studio, MatrixOne is the default underlying foundation and is generally invisible in daily work. For DBAs and data platform roles, it is the primary environment for database instance operations and SQL.

Core capabilities¶

Capability	Description
Hyper-converged engine	One copy of the data and one engine support mixed workloads including transactions, analytics, vectors, and full text, reducing system integration and data movement
AI-native data layer	Vectors and full text are first-class capabilities that directly support RAG, semantic retrieval, and the data retrieval required by agents, rather than an attached parallel data plane
Git-for-Data	Snapshots, time travel, branches, merges, rollbacks, and more make datasets as easy to experiment with, reproduce, and revert as code
Cloud-native elasticity	Separation of storage and compute with workload-based scaling supports a smooth transition from demos to high-concurrency production
MySQL compatibility	Connect through the MySQL protocol and ecosystem tools to reduce migration and learning costs

What users see: database instances¶

When you use MatrixOne directly, your work primarily centers on database instances.

Instance types¶

Type	Description
Serverless	Compute resources scale automatically with the workload; includes free instances (up to five per account, with free quotas, and convertible to production)
Standard instance	Independent, stable node specifications; compute can be billed by subscription or usage, while storage is billed by usage

Create and connect¶

Create: Select a cloud and region, configure capacity and billing policies, set an admin password (at least eight characters containing uppercase and lowercase letters and digits), and configure a network policy (an IP allowlist is recommended), then confirm creation.
Connect: Use MySQL-protocol clients such as Navicat and DBeaver, JDBC, Python, or Go drivers. Connections can use the public network, a private network, or the browser-based Database Management Platform. Obtain connection information from the instance’s Connect entry.
Sign in: admin is the highest-privilege user. For ordinary users, first create users and roles and grant privileges. The sign-in username format is username:role_name.

Operations¶

Creation, connection, monitoring, alerts, events, backup and restore, suspension, and recovery (data is retained for seven days after suspension) are all managed in the Instance Management Platform. Instance monitoring and alerts have the same names as jobs and alerts in an AI Studio workspace, but they are not the same thing and must be viewed in their respective platforms.

Two permission systems: do not confuse them¶

MatrixOne SQL / instance permissions (GRANT, database/table-level permissions, and instance-side accounts and roles) and AI Studio feature-domain permissions are two independent control planes. Both contribute to “Control Your Data,” but their entry points, objects, and granularity differ, and they cannot replace each other. For SQL permissions, see Guide · MatrixOne permissions.

Handoff: the models used above the foundation come from Genesis¶

MatrixOne makes data storable, queryable, governable, and versionable, but it does not perform model inference itself. The models that AI Studio calls while processing data and running agents—chat, embeddings, and workflow model nodes—are provided through the unified model services of Genesis or through the platform’s built-in AI services. The next page introduces this model gateway.