OrionBelt Analytics MCP
An MCP server builds an ontology from a database schema and checks agent SQL before it runs, for PostgreSQL, Snowflake and more
- 58 ★ on the repository
- Last commit
Broad access
Why this level
- The agent runs SQL in the connected database with your credentials
- The database password sits in a .env file next to the server
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/orionbelt-analyticsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent runs SQL in the connected database with your credentials
- The database password sits in a .env file next to the server
This is third-party code. Review the repository files before installing.
What it does
OrionBelt Analytics analyses a database schema and builds an RDF/OWL ontology from it, mapping classes to tables, columns and foreign keys. The agent finds the right tables by the meaning of a question, locates join paths and only then writes the query. Eight databases are supported: PostgreSQL, MySQL, Snowflake, ClickHouse, Dremio, BigQuery, DuckDB with MotherDuck and Databricks SQL. Every query goes through OBQC, which without calling a model checks tables, columns, join conditions, types and groupings against the ontology and catches row-multiplying fan traps in joins. Errors block execution and warnings arrive with the result. There are SPARQL queries, Plotly charts and workspace persistence between sessions.
Who it is for. Data analysts who ask an agent to write SQL against large schemas and want to see errors before the query runs.
Good fit when
- A warehouse schema has hundreds of tables with cryptic names and the agent gets lost in joins
- Aggregation across several one-to-many links gives inflated totals and the cause is not visible
- You need a database ontology in OWL mapped to tables with R2RML
- Questions arrive in another language and schema search must understand them
Not a fit when
- The database is small with ten tables and the agent copes with SQL without an ontology
- You need work with metrics on a ready semantic layer: there is a separate OrionBelt Semantic Layer project for that
- The BUSL-1.1 license does not fit your use case
Example request
Connect to the database, analyse the sales schema, generate the ontology and compute monthly revenue with join checksMCP
- Transport
- http
- Authentication
- not required
Environment variables
POSTGRES_HOST- PostgreSQL host, one of the database setting groups in `.env`.
POSTGRES_PASSWORDsecret- Database user password, other databases use their own variables.
MCP_SERVER_PORT- HTTP server port, 9000 by default.
Limitations
The server runs from source: Python 3.13 or newer and uv are needed and database credentials go into `.env`. The BUSL-1.1 license limits commercial use, so read the terms in the LICENSE file. By default the server listens on `http://localhost:9000` over HTTP and Claude Desktop needs the `mcp-remote` proxy. Schema search uses local ChromaDB embeddings and the first model load takes time. There are 26 tools in all, including SQL execution.
How to disable. Remove the server from the client (`claude mcp remove orionbelt-analytics`), stop the `server.py` process and delete the `.env` with the database credentials.
FAQ
How do I connect the server to Claude Code?
Start `uv run server.py` and run `claude mcp add orionbelt-analytics http://localhost:9000/mcp`.
Which databases are supported?
PostgreSQL, MySQL, Snowflake, ClickHouse, Dremio, BigQuery, DuckDB with MotherDuck and Databricks SQL.
What does OBQC check?
Existence of tables and columns, join conditions against foreign keys, type compatibility, groupings and fan traps. The check does not call a model and errors block execution.
Can I load my own ontology?
Yes, the `load_my_ontology` tool reads a `.ttl` file from the import folder.
What is the license?
BUSL-1.1: the source is open, but commercial use is limited by the terms in the LICENSE file.
Discussion
No questions or reviews yet.
Related
MCP serversThe obsl CLI validates a YAML metrics model and compiles it into SQL for eight dialects, so agents stop hand-writing fact joins.
NVIDIA Auto Ontology semantic layer as an MCP server: SQL gets validated, facts land in a personal knowledge base.
A text-to-SQL MCP server: an agent asks a database in words, gets SQL plus a trace, and a stated refusal when data is missing.
FalkorDB's Text2SQL server: an agent asks the database in plain language and a schema graph helps build correct SQL
An MCP server for a knowledge graph over Unity Catalog tables: the agent picks a domain, reads the ontology and runs GraphQL.
An agent queries GreptimeDB metrics, logs and traces with SQL, TQL and RANGE, with write protection and data masking
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| OrionBelt Analytics (this card) | yes | not required | 58 | |
| OrionBelt Semantic Layer | no | 104 | A semantic layer from YAML models, while this one builds an ontology from a database schema and checks SQL. | |
| DBHub | no | not required | 3,839 | Universal database access, while OrionBelt adds an ontology and query checks. |
README in short
An architecture description, a section on OBQC checks, a quick start from source, connection to Claude Desktop, Claude Code and LibreChat, a table of 26 tools by group and a list of related OrionBelt projects.