How I Use Kiro to Analyze My SQL Server Morning Health Reports

---

Every DBA has a morning routine. Run the health check scripts, scan the output, decide what needs attention before the business day starts. The data collection part is straightforward. The interpretation layer is where time disappears: reading through raw query output across multiple servers, recognizing what is abnormal, deciding what constitutes a finding versus background noise, and producing something coherent enough to bring to change control. That work is slow, repetitive, and depends heavily on one DBA’s pattern recognition on any given morning.

This article covers a workflow that hands the interpretation layer to Kiro, AWS’s agentic IDE. The approach is practical: collect SQL Server health check output in SSMS, save it to a local workspace folder, configure Kiro with a strict evidence-only guardrail, and let Kiro read and analyze all the files at once. The output is a structured technical assessment with severity-ranked findings, confirmed evidence for each one, and a remediation roadmap ready for management and change control review.

This article is written for data professionals who have never used Kiro before. It covers installation, the interface, every setup step, how to get data out of SSMS and into the workspace, how to issue the analysis prompt, and what comes out the other side.

1 What Kiro Is Beginner

Kiro is an agentic IDE built by AWS. It launched in public preview on July 14, 2025 and reached general availability in May 2026. It is the official successor to Amazon Q Developer, which reaches end of support on April 30, 2027.

The most important thing to understand about Kiro is that it is not a chat window. It is an IDE, meaning it opens and works with files on the local computer the same way VS Code does. When a folder is opened in Kiro, Kiro can read every file in that folder. That file awareness is what makes it useful for analyzing health check reports rather than requiring the DBA to paste query output into a chat window line by line.

Kiro is built on Code OSS, the same open-source foundation as Visual Studio Code. VS Code extensions, themes, and keybindings import directly during setup. For DBAs who already use VS Code, the interface is immediately familiar. For those who do not, Section 4 covers exactly what to look for.

Kiro supports SQL as a language natively alongside Python, JavaScript, C#, Go, and others. SQL health check output files, T-SQL scripts, and plain text diagnostic exports are all understood without configuration.

2 Why Kiro Fits SQL Server Health Analysis Beginner

Three Kiro features make it specifically well-suited to this workflow compared to general chat-based AI tools.

Steering Files persist across sessions. A Steering File is a Markdown file stored inside the workspace folder. It acts as a permanent system prompt that applies to every Kiro interaction in that workspace without being re-entered. The evidence-only guardrail rules in Section 6 go into this file. They ensure every analysis follows the same standard regardless of how many days or weeks the workspace is used. A chat window loses this context the moment the session closes.

The workspace reads all files at once. Drop five server health check reports into the input folder, open the workspace in Kiro, and issue a single prompt. Kiro reads every file. Doing the same in a chat tool means pasting five separate files across multiple messages and hoping the tool holds context across all of them.

The output is a saved document, not a conversation. Kiro writes the findings to a file in the workspace. That file becomes a dated record of each morning’s assessment, accumulating over weeks and months into a history that reveals patterns across time. A chat response disappears when the window is closed.

3 Installing Kiro and Signing In Beginner

Download Kiro from kiro.dev. Installers are available for Windows, Mac, and Linux. An AWS account is not required for the free tier or the Pro tier.

Installation steps

1. Go to kiro.dev and click Download
2. Run the installer for the operating system. On Windows this is a standard .exe installer. On Mac it is a .dmg. On Linux a .deb or .rpm package is available depending on the distribution.
3. Launch Kiro after installation completes
4. On the sign-in screen, choose one of the available options: GitHub, Google, AWS Builder ID, or AWS IAM Identity Center. For individual DBAs without an AWS account, GitHub or Google sign-in is the simplest path.
5. After signing in, Kiro opens to the Welcome screen. This is the same layout as VS Code: a sidebar on the left, an editor area in the center, and a panel at the bottom for terminal output.

4 The Kiro Interface: What a DBA Needs to Know Beginner

Four areas of the Kiro interface matter for this workflow.

The Explorer panel (left sidebar). This shows the files and folders in the currently open workspace, identical to VS Code’s file explorer. The Steering File, input reports, and output findings documents all appear here. Click any file to open it in the editor.

The Agent panel. This is where prompts are entered and where Kiro responds. Open it by clicking the Kiro icon in the Activity Bar on the far left, or by pressing Ctrl+L (Windows/Linux) or Cmd+L (Mac). This is where the analysis prompt from Section 12 goes. Kiro reads the workspace files and writes its response into the panel, and can also write files directly to the workspace output folder.

Agent mode settings. At the top of the Agent panel there is a mode selector. Two modes exist: Auto and Review. Auto mode lets Kiro take actions without confirmation. Review mode shows what Kiro is about to do and asks for approval before any file is written or command is run. Always use Review mode for production DBA work. This is the correct setting for a workflow that may produce reports used in change control.

The Steering panel. Kiro shows active Steering Files in the Agent panel sidebar. When a Steering File is present in .kiro/steering/, Kiro displays it as context that is always active. This confirms the evidence-only guardrail is loaded before running the analysis.

5 Creating the Workspace Folder Structure Beginner

Create this folder structure on the local machine before opening Kiro. The folders can be created in Windows Explorer, Mac Finder, or any terminal. The .kiro folder name starts with a dot, which hides it on Mac and Linux by default. It needs to be there for Steering Files and MCP settings to work.

SQL-Health-Assessment/
├── .kiro/
│   ├── steering/
│   │   └── evidence-only.md        -- guardrail: create this in Section 6
│   └── settings/
│       └── mcp.json                -- MCP config: create this in Section 7
├── input/
│   ├── health-check-results/       -- paste exported SSMS output here
│   ├── sql-error-logs/             -- sanitized error log text exports
│   └── supporting-evidence/        -- any additional diagnostic files
├── working/                        -- Kiro uses this as scratch space
├── output/
│   ├── technical-report.md         -- Kiro writes the full findings here
│   └── management-summary.md       -- shorter executive summary
├── README.md
└── .gitignore

After creating the folders, open Kiro and go to File > Open Folder. Select the SQL-Health-Assessment root folder. The Explorer panel on the left now shows the full folder tree. Every health check report file added to the input/health-check-results/ folder will be visible here and readable by Kiro in the next analysis session.

The .gitignore file

If the workspace folder is tracked in version control, create a .gitignore file in the root to prevent health check data and output files from being committed. Only the Steering File, MCP configuration, and README belong in version control.

input/
working/
output/
*.xel
*.bak
*.trn
*.mdf
*.ldf
*.dmp
*.zip
*.7z
.env
*.key
*.pem
credentials*

6 The Evidence-Only Steering File Intermediate

The Steering File is what separates a disciplined AI-assisted analysis from a tool that invents findings. Without it, Kiro brings general SQL Server knowledge to the analysis and may flag things as issues simply because they differ from textbook best practice, regardless of whether the evidence shows an actual problem. The evidence-only Steering File enforces the same standard a senior DBA would apply when writing a findings report.

In the Explorer panel, navigate to .kiro/steering/ and create a new file named evidence-only.md. Paste the following content into it and save.

# Evidence-Only SQL Server Assessment Rules

- Analyze only files located inside this workspace.
- Never claim access to a production SQL Server instance.
- Never request, expose, store, or reuse passwords, tokens,
  connection strings, private keys, or personal data.
- Do not use external websites, external APIs, remote MCP tools,
  or unapproved documentation sources.
- Treat every root-cause statement as unconfirmed unless the
  supplied evidence directly proves it.
- Label conclusions as Confirmed, Suspected, Informational,
  or Not Determinable.
- Cite the exact input filename, section, timestamp, query output,
  or error message supporting each finding.
- Never execute remediation commands.
- Generate proposed commands only as reviewed examples, clearly
  marked "DO NOT RUN WITHOUT CHANGE APPROVAL."
- Do not modify files under the input directory.
- Write generated material only under working or output.
- Before finalizing, identify unsupported assumptions and
  remove or relabel them.

Once this file is saved, open the Agent panel in Kiro. The Steering File appears in the context section at the top of the panel, confirming it is active. Every analysis prompt issued in this workspace now runs under these rules automatically.

7 Disabling MCP Connections Intermediate

Kiro supports MCP (Model Context Protocol) servers that allow the agent to connect to external tools and services. For an isolated health check analysis workspace these connections should be disabled. An MCP connection to a database tool or external service could allow the agent to query live systems, which is exactly what the evidence-only approach is designed to prevent.

In the Explorer panel, navigate to .kiro/settings/ and create a file named mcp.json. Paste the following and save.

{
  "mcpServers": {}
}

This empty configuration disables all MCP server connections for this workspace. If Kiro is managed centrally by an IT administrator, confirm whether MCP is already disabled globally or restricted to an approved list before adding this file.

8 Optional: Stronger Isolation with a Dev Container Advanced

For teams with stricter data handling requirements, running the workspace inside a Docker container isolates the analysis environment from the rest of the computer. This requires Docker Desktop or another company-approved container runtime to be installed.

Create a folder named .devcontainer in the workspace root, then create devcontainer.json inside it with the following content.

{
  "name": "SQL Health Assessment",
  "image": "mcr.microsoft.com/devcontainers/base:ubuntu",
  "workspaceFolder": "/workspace",
  "mounts": [
    "source=${localWorkspaceFolder},target=/workspace,type=bind"
  ],
  "runArgs": [
    "--cap-drop=ALL",
    "--security-opt=no-new-privileges"
  ],
  "remoteUser": "vscode"
}

When Kiro detects this file it offers to reopen the workspace in a container. The workspace files are mounted into the container but the container itself has no elevated privileges and no additional network access beyond what Docker allows by default.

9 Step 1: Export Health Check Data from SSMS Beginner

The source data for Kiro comes from SSMS query output. There are three ways to get that output into a text file. Choose whichever fits the environment and workflow.

Method A: SSMS Results to File (simplest for manual runs)

1. Open SSMS and connect to the SQL Server instance
2. Open the health check script in a new query window
3. Before executing, switch the output mode: on the Query menu, select Results To, then Results to File. Alternatively use the keyboard shortcut Ctrl+Shift+F.
4. Execute the query with F5
5. SSMS immediately opens a Save Results dialog. Navigate to the workspace input/health-check-results/ folder. Name the file with today’s date and the server name, for example SQL01-2026-07-15.txt. Click Save.
6. Repeat for each server being monitored

Method B: sqlcmd (for scripted or automated export)

sqlcmd is the SQL Server command-line utility included with SQL Server and SSMS. Run it from a command prompt or batch file to execute the health check script and write the output to a file automatically.

-- Run from Command Prompt or PowerShell
-- -S: server name  -E: Windows authentication  -i: input script  -o: output file
-- Replace server name, script path, and output path with actual values

sqlcmd -S SQL01 -E -i "C:\sql-health-workspace\scripts\morning-check.sql" ^
       -o "C:\sql-health-workspace\input\health-check-results\SQL01-2026-07-15.txt"

Method C: PowerShell with Invoke-Sqlcmd (for automated multi-server export)

PowerShell with the SqlServer module handles multiple servers in a loop and names each output file automatically with the server name and date.

# PowerShell: export health check from multiple servers
# Requires the SqlServer module: Install-Module SqlServer
# Adjust server list, script path, and output folder

$servers    = @('SQL01', 'SQL02', 'SQL03')
$scriptPath = 'C:\sql-health-workspace\scripts\morning-check.sql'
$outputBase = 'C:\sql-health-workspace\input\health-check-results'
$date       = Get-Date -Format 'yyyy-MM-dd'

foreach ($server in $servers) {
    $outputFile = "$outputBase\$server-$date.txt"
    try {
        Invoke-Sqlcmd -ServerInstance $server `
                      -InputFile $scriptPath `
                      -TrustServerCertificate |
        Format-Table -AutoSize |
        Out-File -FilePath $outputFile -Encoding UTF8
        Write-Host "Saved: $outputFile"
    }
    catch {
        "ERROR connecting to $server`: $_" |
        Out-File -FilePath $outputFile -Encoding UTF8
        Write-Host "Error logged for $server"
    }
}

10 Step 2: Sanitize the Files Before Opening Kiro Intermediate

Before the files go into the workspace, review them for sensitive information. The health check scripts in the SQLYARD toolkit collect metrics and status information, not data values from application tables. If using custom scripts, check the output carefully. Sanitization must happen before the file enters the workspace, not afterward.

Safe to include	Remove or mask before saving
SQL Server version and build number	Usernames and employee names
Wait statistics and performance counters	Passwords, tokens, keys, connection strings
Error numbers, timestamps, and affected components	Customer names and personal information
Database names, job names, AG replica names	Full business data or query result rows
Job history, backup history, storage, and integrity results	Public IPs or hostnames if prohibited by policy
Execution plan details where approved	Stored procedure source code containing proprietary logic

11 Step 3: Open the Workspace in Kiro Beginner

With today’s sanitized report files in the input/health-check-results/ folder, open Kiro.

1. In Kiro, go to File > Open Folder
2. Navigate to and select the SQL-Health-Assessment root folder
3. The Explorer panel on the left shows the full folder tree. Confirm the report files are visible under input/health-check-results/
4. Open the Agent panel by pressing Ctrl+L (Windows/Linux) or Cmd+L (Mac), or click the Kiro icon in the Activity Bar on the far left of the window
5. In the Agent panel, look at the top context area. The evidence-only.md Steering File should appear as active context. If it does not appear, verify the file exists at .kiro/steering/evidence-only.md and reload the window
6. Confirm the agent mode is set to Review, not Auto. The mode selector appears at the top of the Agent panel input area

The workspace is now ready. Kiro has read the folder structure, the Steering File is active, MCP is disabled, and the report files are in place. The analysis prompt can be issued.

12 Step 4: Run the Analysis Prompt Intermediate

Type or paste the following prompt into the Agent panel input box and press Enter. The Steering File is already active so these rules are in addition to the guardrail, not a replacement for it. Adjust the folder paths and date in the first line to match the actual files in the workspace.

Analyze all SQL Server health-check reports in the
input/health-check-results/ folder for today (2026-07-15).
Also read any error log exports in input/sql-error-logs/ if present.

Create a professional technical assessment that can be used
to justify remediation work.

Follow these rules strictly:

1. Use only evidence contained in the provided data.
2. Do not assume a root cause.
3. Do not invent missing configuration details, server history,
   workload information, or business impact.
4. Clearly separate confirmed findings from items that require
   additional investigation.
5. For every finding, quote or reference the exact evidence
   that supports it.
6. If the evidence is insufficient, state:
   "Root cause cannot be confirmed from the provided data."
7. Do not describe something as an issue merely because it differs
   from a general best practice. Explain whether the data shows
   an actual risk, impact, or failure.
8. Rank findings as Critical, High, Medium, Low, or Informational,
   and explain the reason for the severity.
9. Do not recommend making a production change without first
   identifying the validation or testing required.
10. Do not claim that a recommendation will solve the problem
    unless the evidence supports that conclusion.

For each finding provide:
- Finding title and severity
- Status: Confirmed, Suspected, or Informational
- Affected server, database, job, file, query, or component
- Exact supporting evidence (quote or reference the source file
  and the specific line or value)
- Observed condition
- Confirmed impact
- Potential impact, clearly labeled as potential
- Most likely causes, clearly labeled as hypotheses
- Additional evidence needed to confirm the root cause
- Recommended next diagnostic step
- Recommended remediation
- Risk of making the change
- Validation steps after remediation
- Priority and suggested order of work

At the beginning include an executive summary with:
- Number of confirmed issues
- Number of suspected issues
- Highest-risk findings
- Immediate actions
- Items that require more evidence
- Overall health assessment

At the end include a remediation roadmap with:
- Immediate actions
- Actions within 7 days
- Actions within 30 days
- Longer-term improvements

Use professional, factual language suitable for management,
technical leadership, and change-control review.
Avoid exaggerated language. Every conclusion must be traceable
to evidence in the supplied files.

Write the full technical report to output/technical-report.md.
Write a shorter executive summary to output/management-summary.md.

Before completing the report, review every statement and remove
or relabel anything not directly supported by evidence.

After entering the prompt and pressing Enter, Kiro reads the workspace files and begins its analysis. In Review mode it shows each proposed action before taking it, including creating or writing the output files. Approve each action as it appears. The analysis typically completes within one to three minutes depending on the number and size of the report files.

13 Step 5: Challenge the Output and Review the Findings Intermediate

After Kiro writes the output files, the analysis is not complete. One more step is required before the findings are usable for production decisions.

In the Agent panel, issue this follow-up prompt:

Review the technical report you just wrote in output/technical-report.md.
List every statement in that report that is not directly proven
by specific evidence from the files in the input/ folder.
For each such statement, identify whether it should be removed,
relabeled as Suspected rather than Confirmed, or flagged as
requiring additional investigation.

This challenge step forces Kiro to audit its own output with the same evidence standard it was instructed to apply during analysis. It surfaces any findings where the agent defaulted to general SQL Server knowledge rather than actual evidence from the supplied reports. Any statement that fails this check does not belong in a report that goes to change control.

After the challenge, open output/technical-report.md in the Kiro editor and review the findings manually as a qualified DBA. Apply the four-check accuracy gate from Section 14 to each finding before approving the report for distribution.

The ten-step repeatable weekly workflow

1. Create a new dated assessment folder from the approved template if archiving each week separately
2. Export only the required SSMS health-check results and error logs
3. Sanitize the files against the data preparation table in Section 10 before copying them to the input folder
4. Confirm the workspace contains an empty MCP configuration and the evidence-only Steering File is present
5. Open the workspace in Kiro. Confirm the Steering File shows as active context in the Agent panel
6. Ask Kiro to inventory the supplied evidence files before writing any conclusions: “List all files you can see in the input/ folder and summarize what each one contains.”
7. Run the analysis prompt from Section 12
8. Issue the challenge prompt: “List every statement not directly proven by the supplied evidence.”
9. Correct or remove unsupported statements. Have a qualified DBA review all severity ratings, business-impact claims, and remediation steps.
10. Move the approved report to the designated company repository and delete or archive the temporary working copy according to company retention policy

14 Going Further: Higher-Level Tasks Worth Trying at Work Intermediate

The morning health check workflow is a repeatable, structured task and a good starting point. Once the workspace pattern is established, the same approach works for significantly larger analytical tasks where Kiro’s ability to read an entire codebase or file set in one pass provides the most value over a standard chat tool.

The key shift is giving Kiro a high-level objective and letting it determine what to read and how to approach it, rather than issuing line-by-line requests. These are the kinds of tasks where agent-based tools consistently outperform standard AI chat interactions.

SQL Server examples worth trying

-- Paste into the Kiro Agent panel with relevant files in the workspace

-- Architecture and dependency mapping
"Analyze all stored procedures in the input/ folder and explain the overall
architecture. Identify which procedures call each other, what the main data
flows are, and which tables are most central to the application."

-- Security audit
"Find all SQL injection risks in the stored procedures in the input/ folder.
For each risk, identify the exact parameter and the stored procedure it
appears in. Do not flag parameterized queries as risks."

-- Performance investigation
"Explain why the stored procedure in input/usp_GetOrders.sql is likely slow
based on the execution plan in input/usp_GetOrders_plan.xml and the wait
statistics in input/wait-stats-2026-07-15.txt. Suggest specific improvements
and explain the evidence for each one."

-- Refactoring opportunity identification
"Identify duplicated business logic across the stored procedures in the
input/procedures/ folder. List each instance of duplication, explain what
it does, and describe how it could be consolidated."

-- Migration planning
"Create a step-by-step migration plan to move the tables and procedures
described in input/schema-export.sql from the current on-premises SQL Server
configuration to Azure SQL Database. Flag any incompatibilities and list
the features that require changes."

-- Index and query analysis across a database
"Review all query plans in the input/plans/ folder and identify the top five
most expensive operations by estimated cost. For each, describe what the
operator is doing, why it is expensive, and what index or query change
would reduce the cost."

15 The Four-Check Accuracy Gate Intermediate

Before any finding from the report is used to justify a production change, it must pass all four of these checks during the DBA review. Any finding that cannot pass all four is not ready for change control. Label it as requiring additional investigation and do not allow it to drive a production change.

Check	Required Evidence
Condition exists	Exact log entry, query result, configuration value, or repeated measurement from the supplied files
Impact exists	Observed failure, measurable degradation, documented exposure, or affected service. Not a theoretical risk.
Cause is established	Correlated evidence or controlled validation, not timing alone or general best-practice reasoning
Remediation is justified	Change rationale, risk review, test plan, rollback plan, and success criteria are all present

16 IT and Security Verification Checklist Intermediate

Answer all of these questions with the IT or security team before deploying this workflow in any company environment. These are not optional for regulated industries or environments with data governance requirements.

• Is Kiro approved for company source code and diagnostic data?
• Which identity provider and session policy control authentication? Kiro sessions have an eight-hour duration by default.
• Are prompts, file contents, and tool actions retained or audited by AWS?
• Is customer data prohibited even after masking?
• Are MCP servers disabled, centrally allow-listed, or user-configurable?
• Which Kiro and AWS endpoints are permitted through the corporate proxy?
• Is use inside Docker, a dev container, or a company VM supported and approved?
• What is the required retention and deletion policy for generated reports and the temporary workspace files?

17 Kiro Pricing and Plans Beginner

All pricing is verified from kiro.dev as of July 2026. Credit consumption varies by model. Claude Sonnet tasks consume fewer credits than Claude Opus tasks. Kiro selects the model automatically in Auto mode based on task complexity.

Plan	Price	Credits	Best for
Free	$0	50 interactions/month	Evaluating the workflow before committing
Pro	$19/month	1,000 credits/month	Individual DBA running daily or weekly assessments
Pro+	Varies	Higher credit pool	Active multi-server daily analysis
Power	$184/month	10,000 credits/month	Teams running multiple automated agent sessions
Enterprise	Contact AWS	Dedicated	GovCloud, CloudTrail audit logging, IAM Identity Center, admin controls

The Pro plan at $19/month is the right starting point for an individual DBA. A single morning analysis session analyzing three to five server reports typically consumes a small fraction of the monthly 1,000 credit allocation. Monitor credit usage in the Kiro dashboard during the first two weeks before committing long-term.

References

• kiro.dev: Official Kiro IDE download, pricing, and documentation
• kiro.dev/docs: Steering Files, agent modes, MCP configuration, and workspace setup
• Microsoft Docs: SSMS Query Results to Text and Results to File options
• AWS Bedrock: The model infrastructure powering Kiro
• SQLYARD: SQL Server Health Check Toolkit
• SQLYARD: Always On Availability Groups Complete Guide
• SQLYARD: SQL Server Wait Statistics Guide