Turn SQL Server Into a REST API — and an MCP Server for Your AI Agent
Connect Microsoft SQL Server or Azure SQL, choose the tables and views to expose, and publish documented REST endpoints that your AI assistant can also call as tools.
Free plan includes a hosted MCP server — no card required.
SQL Server tends to sit at the centre of businesses that have been running for a while — ERP data, invoicing, stock, customers. It is also where integration requests pile up, because every new tool, dashboard or partner wants a slice of the same tables and nobody wants to hand out database credentials to do it.
Dataloom turns that database into a governed HTTP surface. Connect with your server address, database, username and password, pick the tables and views worth exposing, and publish REST endpoints with interactive documentation and an OpenAPI description generated from your live schema.
Because Dataloom queries your server directly at request time, there is no synchronisation to manage and no second copy of the data to secure. Schema-qualified objects work as you would expect, so dbo.Invoices and reporting.CustomerActivity are both first-class.
Connecting SQL Server or Azure SQL
Add a data source with your server host, port (1433 by default), database name and SQL credentials. Encryption is always on for the connection, which is what Azure SQL requires; self-hosted instances with self-signed certificates are handled too. Dataloom tests the connection before saving.
Dataloom then reads your schema and lists the available tables and views along with their columns, types and keys. Views are frequently the best thing to publish here: you can express the joins and business rules in SQL once, then expose the result as a single tidy endpoint rather than making callers reassemble it.
For Azure SQL, remember that the server firewall must allow the connection. A dedicated login with rights to just the objects you intend to publish is the cleanest setup, and it keeps the blast radius of the API to exactly what you chose.
What you get with SQL Server
- Tables and views, including schema-qualified names like dbo.Invoices
- Works with on-premises SQL Server and Azure SQL Database
- Encrypted connections by default
- Filtering, sorting and pagination on list endpoints
- Custom SQL endpoints with named :parameter binding (Starter and above)
- Bracket-quoted identifiers, so reserved words are safe
An example endpoint
Customers with an overdue balance, worst first.
GET /finance/customeractivity
?filter={"Province":"Gauteng","OverdueRatio":{"gte":1}}
&sort=OverdueRatio&order=desc
&page=1&pageSize=10
X-API-Key: your_api_key{
"data": [
{ "CustomerId": 31, "CustomerName": "Harbour Supplies",
"Province": "Gauteng", "OverdueRatio": 38.98 },
{ "CustomerId": 2, "CustomerName": "Salty Forest",
"Province": "Gauteng", "OverdueRatio": 35.96 }
],
"meta": { "page": 1, "pageSize": 10, "totalCount": 12,
"totalPages": 2, "hasNextPage": true }
}Filtering, sorting and pagination work the same way on every endpoint, whatever the source. See the no-code API builder for how endpoints are put together.
Connect Claude to SQL Server
Each endpoint you publish becomes a tool on your workspace’s hosted MCP server automatically. For SQL Server shops this is often the fastest route to a useful internal AI assistant, because the reporting views you already maintain become answerable questions without any new plumbing.
Ask for "the ten customers with the worst overdue ratio in Gauteng" and the assistant calls your published endpoint with a real filter and sort, then reads the result. The tool schema carries your actual column names, so it filters on OverdueRatio because that column exists, not because it sounded likely.
Everything stays inside the boundary you set. Unpublished tables are invisible, the connection is authorised with OAuth 2.1 rather than a shared secret, and any assistant can be revoked on its own from your workspace.
// mcp-config.json
{
"mcpServers": {
"dataloom": {
"url": "https://mcp.dataloom.cloud/w/your-workspace",
"transport": "http"
}
}
}Read more about the hosted MCP server, or follow the step-by-step setup guide.
SQL Server questions, answered
Does this work with Azure SQL Database?
Yes. Azure SQL is fully supported — connections are encrypted by default, which is what Azure requires. Make sure the Azure SQL server firewall permits the connection.
Can I publish a view instead of a raw table?
Yes, and it is usually the better choice. Put the joins and filtering logic in the view, then publish the view as one endpoint so callers get a clean, stable shape.
How are stored-procedure-style queries handled?
On the Starter plan and above you can create custom SQL endpoints and bind values with named :parameters, which covers most cases where you would otherwise reach for a procedure.
Is Windows authentication supported?
Connections use SQL authentication with a username and password. Create a dedicated SQL login scoped to the objects you plan to publish.