Sidemantic
sidemantic.com
- Free plan
- Web
- API
- Self-hosted
What Sidemantic is
Sidemantic is an open-source semantic layer for defining metrics once and querying them consistently across tools. It runs on top of a data warehouse rather than acting as an ETL or BI tool. Models can be written in YAML, SQL, or Python, or imported from formats including Cube, MetricFlow, LookML, Malloy, Snowflake Cortex, and others. Native connections include DuckDB, MotherDuck, PostgreSQL, BigQuery, Snowflake, ClickHouse, Databricks, and Spark SQL, with additional databases available through ADBC. Queries work through semantic SQL, a Python API, CLI, MCP server, or the Postgres wire protocol. Declared relationships let cross-model queries infer join paths, while symmetric aggregates help prevent double-counting in one-to-many joins. MCP exposes metrics to LLMs with described guardrails. An interactive workbench includes demo data without database setup. The project is AGPL-3.0 and self-hosted.
Who it suits
Sidemantic suits teams that need governed metrics across warehouses and query interfaces. It is also aimed at users connecting metrics to LLMs through MCP.
Good to know, and what to watch
- Open source under AGPL-3.0
- Models authored in YAML, SQL, or Python
- Many warehouse adapters
- Multiple query interfaces
- Workbench runs with built-in demo data
- Not a BI or ETL tool
- Project describes itself as young
- Format conversion may have rough patches
- Postgres protocol server may have rough patches
The verdict
Sidemantic centralizes metric definitions and exposes them through several interfaces and warehouse connections. Its young-project caveat matters when relying on format conversion or the Postgres protocol server.
What the maker publishes
- Aggregation correctness
- Symmetric aggregates are used to prevent double-counting across one-to-many joins. source
- AI access
- Sidemantic exposes metrics to LLMs through MCP and describes built-in guardrails for governed AI access. source
- AI access
- Sidemantic can expose metrics to LLMs through MCP, with built-in guardrails for governed access. source
- Format imports
- It imports semantic models from formats including Cube, dbt MetricFlow, LookML, Hex, Rill, Superset, Omni, Malloy, Snowflake Cortex, OSI, AtScale, GoodData, Holistics, and ThoughtSpot. source
- Interactive demo
- The site offers an interactive workbench with built-in demo data that does not require a database. source
- Join handling
- Cross-model queries can infer join paths from declared relationships, and symmetric aggregates are designed to prevent double-counting across one-to-many joins. source
- Join inference
- Cross-model queries infer join paths from declared relationships without requiring a JOIN keyword in semantic SQL. source
- License
- The GitHub repository identifies the project as open source under the AGPL-3.0 license. source
- Maker
- The site footer identifies Sidequery Corporation as the company. source
- Maturity
- The project describes itself as young and notes that format conversion and its Postgres protocol server may have rough patches. source
- Model authoring
- Models can be authored in YAML, SQL, or Python. source
- Model authoring
- You can define semantic models in YAML, SQL DDL, or Python. source
- Model imports
- It imports semantic models from more than 15 formats, including Cube, MetricFlow, LookML, Malloy, and Snowflake Cortex. source
- Not a BI or ETL tool
- Sidemantic runs on top of a warehouse and is described as neither an ETL tool nor a BI tool. source
- Pre-aggregation limit
- Pre-aggregation routing is disabled by default and must be explicitly enabled. source
- Purpose
- Sidemantic is an open-source semantic layer for defining metrics once and querying them consistently across tools. source
- Purpose
- Sidemantic is a semantic layer for defining metrics once and querying them consistently across tools. source
- Query interfaces
- Users can query through semantic SQL, Python API, CLI, MCP server, or the Postgres wire protocol. source
- Query interfaces
- Sidemantic provides a SQL interface, Python API, CLI, MCP server, and Postgres wire protocol. source
- Support
- The maker links to documentation and a Discord community for the project. source
- Warehouse connections
- Documented database connections include DuckDB, MotherDuck, PostgreSQL, BigQuery, Snowflake, ClickHouse, Databricks, and Spark SQL. source
- Warehouse support
- Native adapters are listed for Snowflake, BigQuery, Postgres, ClickHouse, Databricks, and Spark, with additional databases supported via ADBC. source
- Workbench
- The workbench can run with built-in demo data without a database setup. source
Checked against the maker's pages on 9 Oct 2026.