What Is Microsoft Fabric Data Agent?

Fabric Data Agent is a generally available feature within Microsoft Fabric that lets users ask natural-language questions over data stored in Fabric OneLake. It connects to lakehouses, warehouses, Power BI semantic models, and KQL databases, then generates SQL, DAX, or KQL queries to return answers. Users can add custom instructions, and example queries to guide the agent. It integrates with Microsoft Purview for governance and can surface answers through Microsoft 365 Copilot, Teams, or Copilot Studio.

Where Fabric Data Agent Falls Short

• Fabric only. The data agent queries data stored in Fabric OneLake. It cannot connect to external databases like Oracle, PostgreSQL, MySQL, or on-premises SQL Server instances outside of Fabric. If your data estate spans multiple platforms, the agent only covers the Fabric portion.
• Cloud only. Fabric is a cloud-native SaaS platform with no on-premises deployment option. For regulated industries (banking, healthcare, government) that require data to remain within their own perimeter, this is a fundamental constraint. Cross-region queries are also restricted: the data source and the agent must be on the same capacity region.
• Locked LLM. The agent uses Azure OpenAI under the hood. You cannot change the LLM, bring your own model, or point to a self-hosted inference endpoint. This limits control over model selection, cost, and data flow.
• Manual context only. The agent has no automated semantic layer. Context is limited to developer-provided instructions and up to 100 example query pairs per data source. Someone must write and maintain these as data changes — there is no automatic learning from query history or BI tools.

How Tursio Is Different

• Platform-agnostic. Tursio connects to SQL Server, Oracle, PostgreSQL, Snowflake, Databricks, MySQL, and flat files (CSV, Parquet). One search layer across all your databases.
• On-premises first. Tursio deploys as a single Docker image inside your environment. No data leaves your perimeter. This is not a cloud-only option with an on-prem afterthought; it is the primary deployment model.
• Bring your own LLM. Tursio supports bring-your-own-LLM keys (OpenAI, Anthropic, Azure OpenAI) and can also point to self-hosted models (e.g., vLLM, Ollama). You control which model runs inference, where it runs, and what it costs.
• Operator trees, not just text-to-SQL. Tursio does not simply translate natural language into a SQL string. Its planner decomposes questions into a logical operator tree, which it then compiles into optimized, multi-step query plans. This is architecturally deeper than one-shot NL2SQL and produces more reliable results on complex, multi-table questions.
• Automated context graph. Rather than relying on manually written instructions and up to 100 example queries, Tursio automatically builds a persistent context graph from schema, query history, and BI tool definitions (Power BI, Tableau). Join paths, column semantics, dimensions/measures, and business aliases are inferred and maintained automatically.

Side-by-Side Comparison

Bottom Line

Fabric Data Agent is a convenient feature for organizations already committed to Microsoft Fabric. It works well for simple conversational queries over Fabric data, especially when paired with Power BI semantic models and Purview governance. But it is Fabric-only, cloud-only, locked to Azure OpenAI, capped at 25 rows, and limited to one-shot text-to-SQL generation with manually curated context.

Tursio is a different category of product: a platform-agnostic, on-premises enterprise search layer that builds a persistent context graph across all your databases, decomposes questions into optimized operator trees, and returns full, deterministic results. If your data lives across multiple platforms, or you need on-premises deployment, LLM flexibility, or production-grade query planning, Tursio is the right choice.