GitHub Copilot Agent Mode is now available as a preview feature in SQL Server Management Studio (SSMS) 22.7, shipped June 9, 2026, introducing autonomous multi-step task automation for database administrators alongside traditional chat-based assistance. This is a genuinely different capability from the earlier Ask mode that reached general availability in March 2026, which focused on answering questions and drafting SQL queries. Agent mode operates independently across complex workflows, requesting approval before running each query or command.

What Makes This Different From Regular Copilot Chat?

The distinction matters because Agent mode in SSMS is built on a local Model Context Protocol (MCP) server named sql-tools, with a predefined set of tools specifically designed for database work. This is the same MCP protocol already documented in depth elsewhere, but here Microsoft implemented it as a first-party client rather than relying on community-built servers. The preview version defaults to read-only mode, meaning every proposed action requires explicit approval before execution, which addresses a key concern for database administrators managing production systems.

Three specific capabilities make Agent mode directly relevant to DBA work rather than just general developer productivity. First, it runs on that local MCP server architecture. Second, it ships with DBA-specific built-in skills. Third, it defaults to read-only in this preview phase. These three elements together create a tool designed from the ground up for database administration tasks, not retrofitted from developer-focused features.

What Built-In Skills Does It Include?

The built-in skills cover core DBA responsibilities: troubleshooting performance issues, analyzing missing indexes, and reviewing configuration settings. Beyond these predefined capabilities, administrators can create custom skills for additional workflows specific to their environment. This flexibility means the tool can adapt to organizational needs without requiring constant updates from Microsoft.

The governance model introduces a configurable execution context controlled through a CONSTITUTION.md file at the database level. When an agentExecuteAsUser value is specified in that file, all queries in both Ask and Agent mode execute under the context of that specified user or login rather than the connecting account's own permissions. This separation is crucial for audit trails and compliance requirements in regulated industries.

How Does Security Actually Work in Agent Mode?

Microsoft clearly states that the actual security boundary is SQL Server's own permission enforcement, not Copilot's approval system. The approval prompt functions as a workflow control, not as a substitute for correctly scoped SQL Server permissions on whatever account Agent mode is executing under. This distinction prevents a false sense of security where administrators might assume the approval step provides protection it does not actually deliver.

To use Agent mode, administrators need SSMS 22.7 or later with the AI Assistance workload installed, plus a GitHub account with Copilot access. Notably, Microsoft's documentation indicates GitHub Copilot can be used free of charge in SSMS, separate from a paid GitHub Copilot subscription used elsewhere. This pricing structure removes a potential barrier to adoption for organizations evaluating the tool.

How Agent Mode Fits Alongside Other Copilot Products

Microsoft uses the word "Copilot" for at least four distinct products, and understanding which tool serves which purpose prevents confusion during implementation. Agent mode in SSMS sits alongside but does not compete with Copilot Studio, which is a low-code platform for building custom AI agents that other people, business users, executives, and support staff use through Teams or chat channels. Someone could reasonably use both: an SSMS Agent mode session for personal, hands-on diagnostic work, and a Copilot Studio agent published for a wider team that should never need direct SSMS access at all.

The other Copilot products include Microsoft 365 Copilot, the general AI assistant embedded across Word, Excel, Teams, and Outlook; GitHub Copilot Agent Mode in VS Code via the MSSQL extension, a developer-focused agent that connects to SQL Server and retrieves schema; and Copilot in SSMS itself, which as of SSMS 22.7 includes the genuine Agent Mode being discussed here. Each serves a different audience and use case.

Steps to Implement GitHub Copilot Agent Mode in Your Environment

• Prerequisites: Install SSMS 22.7 or later with the AI Assistance workload, and ensure you have a GitHub account with Copilot access. No paid subscription is required for SSMS usage.
• Configure Execution Context: Create or update a CONSTITUTION.md file at the database level to specify the agentExecuteAsUser value, controlling which SQL Server login or user account the agent executes queries under.
• Set Approval Workflows: Establish approval processes for Agent mode actions in your organization, understanding that approval prompts are workflow controls, not security boundaries, and that SQL Server permissions remain the actual security enforcement mechanism.
• Test with Read-Only Tasks: Begin with read-only diagnostic tasks like performance troubleshooting and missing index analysis before expanding to custom skills that might modify database configurations.
• Define Custom Skills: Identify DBA workflows specific to your environment that go beyond the built-in skills for troubleshooting, index analysis, and configuration review, then create custom skills to automate those tasks.

The June 2026 release of Agent mode in SSMS represents a meaningful shift in how database administrators can interact with SQL Server. Rather than manually executing queries or relying on chat-based assistance, DBAs can now delegate multi-step diagnostic and maintenance tasks to an autonomous agent that understands database-specific operations and defaults to safety through read-only execution and approval requirements. The tool's integration with existing SQL Server permission models and its free availability in SSMS removes both technical and financial barriers to adoption.

" }