Redigeeri

How GitHub Copilot works with the MSSQL extension

The GitHub Copilot integration with the MSSQL extension for Visual Studio Code exposes several surfaces where artificial intelligence (AI) assists your SQL development. Each surface has different capabilities, different access to your database schema, and different ideal use cases. This article explains which surface handles which scenario, so you can choose the right tool for each task.

Key takeaways

  • The @mssql chat participant, agent mode tools, Schema Designer, and Data API builder all have schema awareness when connected to a database.
  • Inline completions (ghost text while typing in a .sql file) come from GitHub Copilot's model directly and don't see your connected database schema.
  • For schema-aware suggestions while writing SQL, use the @mssql chat participant in chat instead of inline ghost text.

GitHub Copilot surfaces at a glance

Surface Provided by Schema-aware? Best for
Chat participant (@mssql) MSSQL extension Yes (connected database) Questions, explanations, query authoring
Agent mode tools MSSQL extension contributes tools Yes (via tool calls) Multi-step workflows, delegated changes
Plan mode Visual Studio Code Yes (via @mssql context) Reasoning before writing SQL data definition language (DDL)
Slash commands MSSQL extension Yes Structured prompts for common tasks
Inline completions GitHub Copilot model directly No Generic ghost text while typing
Schema Designer with GitHub Copilot Embedded in MSSQL Schema Designer Yes Visual schema design with AI assistance
Data API builder with GitHub Copilot Embedded in MSSQL Data API builder Yes Entity configuration for REST, GraphQL, and Model Context Protocol (MCP) endpoints

The chat participant: @mssql

The @mssql chat participant is contributed by the MSSQL extension. When you type @mssql in the GitHub Copilot Chat view, the extension receives your prompt along with metadata about your active database connection. The extension can:

  • Read schema information (tables, columns, relationships, stored procedures)
  • Include schema context in the prompt sent to the model
  • Return responses that reference real objects in your database

The chat participant is the primary way to get schema-aware AI assistance. It powers ask mode and edit mode interactions when you address @mssql in chat.

For a scenario-driven walkthrough, see Quickstart: Chat with the @mssql participant (ask mode).

Agent mode tools

Agent mode lets GitHub Copilot autonomously plan and execute work. The MSSQL extension contributes a set of tools (for example, connect, list_databases, run_query) that agent mode can call on your behalf, always with your approval.

Unlike the chat participant (which needs an explicit @mssql mention), agent mode picks up the MSSQL extension's tools automatically when the extension is active. You write natural-language prompts like "Connect to my LocalDev profile and show me the tables in AdventureWorks," and agent mode chooses which tools to invoke.

Agent mode is schema-aware through its tools. Each tool call returns real data from your connected database.

For the full tool reference and example prompts, see Quickstart: Use GitHub Copilot agent mode.

Plan mode

Plan mode is a Visual Studio Code feature that lets GitHub Copilot reason through a request without making changes. When you switch chat to plan mode, GitHub Copilot produces a written plan (often saved as plan.md) that you can review before handing off to agent mode or Schema Designer for execution.

Plan mode is a strong fit for database design. A natural-language product requirements document becomes a reasoned data model that includes tables, junction tables, foreign key direction, and constraints before any SQL DDL is written.

For a walkthrough that pairs plan mode with a product requirements document (PRD), see Quickstart: Use plan mode for spec-driven database design.

Inline completions

Inline completions are the ghost text that appears as you type in the editor. In .sql files, these suggestions come from GitHub Copilot's model directly.

Inline completions don't see your connected database schema. Visual Studio Code's inline completion application programming interface (API) is effectively single-provider: when GitHub Copilot is enabled, third-party extensions (including the MSSQL extension) can't contribute schema-aware ghost text. For schema-aware SQL suggestions, use the @mssql chat participant instead.

This is a platform behavior, not an MSSQL extension limitation. It applies to every third-party extension that wants to contribute inline SQL completions. For a deeper look at what inline completions do offer, see Quickstart: Use inline GitHub Copilot completions in SQL files.

Schema Designer with embedded GitHub Copilot

The MSSQL extension's Schema Designer has GitHub Copilot embedded directly in its canvas. This is a separate pipeline from chat and agent mode. The Schema Designer has its own prompt infrastructure and its own way of applying schema context. You can ask GitHub Copilot to create tables from selected code, add relationships, generate test data, or import external artifacts, and watch the visual diagram update live.

Schema Designer with GitHub Copilot is schema-aware because it operates directly on your connected database.

For scenario-driven examples, see Quickstart: Design schemas visually with embedded GitHub Copilot scenarios.

Data API builder with embedded GitHub Copilot

The Data API builder canvas also embeds a GitHub Copilot chat surface. You can use natural language to configure entities, permissions, and output types (REST, GraphQL, MCP). Like Schema Designer, this surface is schema-aware and uses its own prompt pipeline.

For the full workflow, see Data API builder.

Custom instructions apply across surfaces

Regardless of which surface you use, GitHub Copilot applies any custom instructions you've authored for your project. Instructions files (.github/instructions/*.instructions.md) teach GitHub Copilot your team's conventions, and they're injected into every request whose applyTo glob matches.

This means your naming conventions, file templates, and data type preferences influence ask mode, agent mode, plan mode, and inline completions alike. For how to set this up, see Quickstart: Use custom instructions to align GitHub Copilot with your T-SQL conventions.

Privacy and data handling

Every surface routes through GitHub Copilot's privacy-preserving proxy. Prompts and completions aren't stored and aren't used to train the models. For details, see Limitations and known issues and the GitHub Copilot Trust Center.

Share your experience

To help us refine and improve GitHub Copilot for the MSSQL extension, use the following GitHub issue template to submit your feedback: GitHub Copilot Feedback

When submitting feedback, consider including:

  • Scenarios tested: Let us know which areas you focused on, for example, schema creation, query generation, security, localization.

  • What worked well: Describe any experiences that felt smooth, helpful, or exceeded your expectations.

  • Issues or bugs: Include any problems, inconsistencies, or confusing behaviors. Screenshots or screen recordings are especially helpful.

  • Suggestions for improvement: Share ideas for improving usability, expanding coverage, or enhancing GitHub Copilot's responses.

Related content

  • Last updated on