Publish your Microsoft Security Copilot agent to the Microsoft Security Store

Package and publish your Microsoft Security Copilot agent to the Microsoft Security Store. After building and testing your agent, you'll create a deployment package, configure your offer in Microsoft Partner Center, and submit it for review and certification.

This article is part of a series that walks you through ingesting test data to the data lake, developing an agent and publishing the example IdentityDrift platform solution. The solution includes a Security Copilot agent that correlates identity, endpoint, and cloud security signals to automate threat investigations.

Prerequisites

A working Security Copilot agent
Microsoft Partner Center account

Publishing overview

Publishing process follows the following flow:

Agent Development
Create Deployment Package (.zip)
Create SaaS Offer in Partner Center
Configure Offer Metadata & Listing
Review and publish
Live in Security Store

Create the deployment package

Create a deployment package for your agent that includes the required manifest files.

Prepare your package structure

Create a folder structure for your agent with the following layout:

agent-package/
├── PackageManifest.yaml (required)
└── YourAgentName/
    └── AgentManifest.yaml (required)

Create PackageManifest.yaml

The manifest file describes your package structure and contents. Create PackageManifest.yaml in the root:

manifest:
  - id: "IdentityDriftInvestigationAgent"
    description: "Agent to investigate Identity Threats"
    type: CopilotAgent
schema:
  version: "1.0.0"

Key fields:

id: Name of your Security Copilot Agent (with no spaces)
type: CopilotAgent for Security Copilot agents (other types: SentinelLake (for notebooks))

Download agent manifest

The AgentManifest.yaml is exported from Microsoft Security Copilot after building and running your agent. Download it from Security Copilot under the agent Build tab and select the agent.

Select to expand: IdentityDrift-Investigation- Agent `AgentManifest.yaml` file.

Descriptor:
  Name: IdentityDrift-Investigation-Agent
  Description: >-
    Investigates identity risk by correlating authentication signals, access
    telemetry, endpoint activity, and security alerts
  DisplayName: IdentityDrift-Investigation-Agent
  CatalogScope: UserWorkspace
  Enabled: true
  Prerequisites:
    - MCP.Sentinel
  Icon: ''
SkillGroups:
  - Format: Agent
    Skills:
      - Name: IdentityDrift-Investigation-Agent
        DisplayName: IdentityDrift-Investigation-Agent
        Description: >-
          Investigates identity risk by correlating authentication signals,
          access telemetry, endpoint activity, and security alerts
        Inputs:
          - Name: UserPrincipalName
            Description: User Principal Name to investigate
            DefaultValue: ''
            Required: true
        Settings:
          Instructions: >-
            1. Accept User Principal Name (UPN) Input
            Accept a User Principal Name (UPN) or identifying string as input
            from the prompt. Ensure to use that UPN throughout the analysis.
            2. Global Query Rule (MANDATORY)
            Every query MUST filter to the last 24 hours:
            | where TimeGenerated > ago(24h)
            Never use 7 days, 30 days, or "all time." Always 24h. To avoid
            oversized responses, summarize and limit outputs (do not return raw
            event dumps).
            3. Query Data Lake for CommonSecurity_ID_KQL_CL
            IMPORTANT:
            - Do NOT assume the existence of any specific columns such as
            Action, EventType, or Application
            - Use only columns that exist in the query result
            - Prefer the following safe fields when available:
              - TimeGenerated
              - SourceUserName
              - SourceIP
              - DestinationHostName
              - AdditionalExtensions
              - DeviceCustomString1

            Search CommonSecurity_ID_KQL_CL table records for events that match
            the provided user input (use SourceUserName as the identifier).

            Sample KQL Query (replace {{UserPrincipalName}}):

            CommonSecurity_ID_KQL_CL

            | where TimeGenerated > ago(24h)
                and SourceUserName has '{{UserPrincipalName}}'
            | summarize
                TotalEvents=count(),
                MFA_Approved=countif(DeviceCustomString1 has "Approved"),
                PrivilegedActions=countif(DeviceCustomString1 has "Privilege"),
                SensitiveAccess=countif(DeviceCustomString1 has "Sensitive"),
                Activities=makeset(AdditionalExtensions),
                TargetResources=makeset(DestinationHostName),
                IPs=makeset(SourceIP)
                by SourceUserName

            4. Query Data Lake SigninLogs_KQL_CL Table
            - Same user input
            - Filter last 24 hours
            - Extract:
              - Sign-in success vs failure
              - IP diversity
              - Result descriptions

            5. Query Data Lake AADRiskyUsers_KQL_CL Table

            - Same user input
            - Filter last 24 hours
            - Extract:
              - RiskLevel
              - RiskState
              - RiskLastUpdatedDateTime

            6. Query Data Lake DeviceProcessEvents_KQL_CL Table
            - Same user input
            - Filter last 24 hours

            - Identify suspicious post-authentication activity

            Guidance:
            - Remove domain from UPN to derive AccountName
            - Look for LOLBins in FileName column 
              - powershell.exe
              - cmd.exe
              - kubectl.exe
              - az.exe

            7. Query Microsoft Defender for Cloud SecurityAlert Table

            Query SecurityAlert to identify confirmed runtime threats related to
            Kubernetes or cloud workloads that may correlate with identity
            activity.

            - Alerts generated by Microsoft Defender for Cloud
            Kubernetes‑related alert types such as:
              - K8S.NODE_MalwareBlocked
              - K8S.NODE_DriftBlocked

            Guidance:

            - Filter to last 24 hours

            - Do NOT expect user identity fields in SecurityAlert

            - Extract:
              - AlertType
              - AlertSeverity
              - CompromisedEntity (ClusterName)
              - Context from ExtendedProperties

            8. Correlation & Reasoning

            Use the Sentinel Data Exploration MCP tool to correlate activity
            between CommonSecurity_ID_KQL_CL , SigninLogs_KQL_CL,
            AADRiskyUsers_KQL_CL, DeviceProcessEvents_KQL_CL and
            SecurityAlert_KQL_CL

            Match overlapping:
            - User identifiers

            - IP addresses

            - Device names

            - Authentication privilege escalation

            - Suspicious endpoint execution post authentication compromise

            9. Surface Key Insights

            Identify:

            - Risky sign-ins followed by privileged access

            - Unexpected MFA approvals

            - Access to vulnerable or high-value workloads

            - Privilege escalation preceding endpoint activity and Kubernetes
            control‑plane actions

            - Suspicious endpoint or Kubernetes tooling execution

            - Defender for Cloud alerts occurring after identity or
            control‑plane activity


            10. Provide Summary Findings

            Summarize:

            - MFA outcomes

            - Sign-in success vs failure trends

            - Identity risk posture

            - Privileged access highlights

            - Endpoint execution signals

            - Defender for Cloud security alerts and their timing


            Highlight discrepancies or noteworthy observations across identity,
            access, and endpoint telemetry.


            ### Sample Automation Flow (Short Version)

            1. Query **CommonSecurity_ID_KQL_CL** for identity access context

            2. Query **SigninLogs_KQL_CL** and **AADRiskyUsers_KQL_CL** for
            authentication and risk posture

            3. Query **DeviceProcessEvents_KQL_CL** for endpoint behavior

            4. Query **SecurityAlert_KQL_CL** for Defender for Cloud runtime
            threats

            5. Correlate all signals using Sentinel MCP and surface actionable
            security insights
        ChildSkills:
          - list_sentinel_workspaces
          - search_tables
          - query_lake
AgentDefinitions:
  - Name: IdentityDrift-Investigation-Agent
    DisplayName: IdentityDrift-Investigation-Agent
    Description: >-
      Investigates identity risk by correlating authentication signals, access
      telemetry, endpoint activity, and security alerts
    Product: IdentityDrift
    Publisher: IdentityDrift
    Settings:
      - Name: UserPrincipalName
        Description: User Principal Name to investigate
        Required: true
    Triggers:
      - Name: DefaultTrigger
        DefaultPollPeriodSeconds: 0
        ProcessSkill: IdentityDrift-Investigation-Agent.IdentityDrift-Investigation-Agent
    RequiredSkillsets:
      - MCP.Sentinel
      - IdentityDrift-Investigation-Agent
    PreviewState: Private
    PublisherSource: Custom
    AgentSingleInstanceConstraint: None

Best practices for AgentManifest.yaml

The Security Store review team runs a thorough validation on AgentManifest.yaml before approving offer. The following are the most common failure points observed:

The product and publisher must be the ISV name, not generic values. The product and publisher fields under AgentDefinitions must reflect the ISV's actual product name and company name. Do NOT leave them as default values like "Custom" after downloading AgentManifest.yaml from Security Copilot.
```
# Incorrect - do not use generic placeholders
AgentDefinitions:
   Product: Custom
   Publisher: Custom

# Correct - use actual ISV product and publisher name
AgentDefinitions:
   Product: Contoso ThreatOps
   Publisher: Contoso Inc.
```
Settings key names must exactly match skill input names, case-sensitive with no spaces. If your Skill declares an input named UserPrincipalName, the Settings section under AgentDefinitions must use exactly UserPrincipalName. Don't use User Principal Name or userprincipalname or other variation.
```
# Incorrect — Spaces in input names are not allowed.
Settings:
  - Name: User Principal Name

# Correct — Matches input key name exactly. No spaces, case-sensitive.
Settings:
  - Name: UserPrincipalName
```
Input fields must include a Description property. Every input/setting field must have a meaningful Description that helps the user understand what value to provide. Without descriptions, users hovering over input fields in Security Copilot see nothing.
Skill names must be descriptive and not version labels or random characters. When defining Custom Skills in Security Copilot Agent, skill names like "Agent v3", or "Skill_01" are rejected because they do not describe what the skill does. Name skills after their action and target, for example,"GetSignInLogsForUser", "QueryRiskyUsersTable", or "CorrelateEndpointActivity".
RequiredSkillsets must include all integrated Microsoft products. If your agent integrates with Microsoft Sentinel Data Exploration or other Sentinel MCP tools, add MCP.Sentinel to RequiredSkillsets. This is important not just for standards alignment. When MCP.Sentinel is listed, Sentinel visibly appears under the Plugins section in the agent run view.
```
RequiredSkillsets: 
- MCP.Sentinel
```

Don't hardcode time windows in KQL — use input parameters. KQL queries with hardcoded values such as ago(7d) are flagged as inflexible. Replace them with input parameters so the time window is configurable.

# Incorrect — hardcoded time window, inflexible for users and reviewers.
Template: >-
  SigninLogs | where TimeGenerated > ago(7d) ...

#  Correct — parameterized time window.
Template: >-
  SigninLogs | where TimeGenerated > ago({{TimeRange}}) ...

Grammar and spelling audit. The review team checks every text field: Descriptor.Description, SkillGroups skill descriptions, input descriptions, and DisplayName fields. Grammatical errors, for example using redundant phrasing like "for a specific given username", or unnecessary capitalization, are flagged. Run a full grammar check on all YAML text fields before packaging.

Create the ZIP package

For Windows/Linux:

cd agent-package
zip -r agent-package.zip .

For Mac, avoid hidden files:

cd /path/to/agent-package
zip -r agent-package.zip . -x ".*" -x "__MACOSX"

This command creates agent-package.zip while excluding:

Hidden files (starting with .)
macOS system folders (__MACOSX)

Verify package contents:

unzip -l agent-package.zip

You should see:

PackageManifest.yaml
YourAgentName/
YourAgentName/AgentManifest.yaml

Prepare for Partner Center publication

Partner Center is the Microsoft commercial marketplace where you create your offer listing for your Security Copilot agent. Gather the following information and assets:

Agent name and version. The name must not contain Microsoft product names.
Agent description of 1-2 sentences describing the agent's purpose and value proposition.
Agent tasks - a list of what the agent does.
Agent workflow — explicit inputs and outputs with data sources and table names.
Marketing or product page URL for the Links section.
A PDF user guide describing how to install and use the agent from the Security Store.
ISV Logo (216×216 px) and Agent Screenshots (1280×720 px) showing full agent execution and results.
Webhook URL for order notifications.
Pricing model - free or paid.
SCU consumption estimate.

Prepare the agent description before starting in Partner Center

The Security Store review team requires the offer description to include a structured format covering Agent Tasks, Inputs, and Outputs. The expected format is as follows:

[Agent name] is a security investigation agent that integrates with Microsoft Sentinel to [brief purpose statement].

Agent Tasks:
 - Task 1 (e.g., Identity threat triage)
 - Task 2 (e.g., Authentication analysis)
 - Task 3 (e.g., Cross-telemetry correlation and anomaly detection)

Agent Workflow:
Input:
 - UserPrincipalName (UPN) — the user account to investigate
 - Access to Microsoft Sentinel data lake tables (TableA_CL, TableB_CL, ...)
 - Time range used for queries: TimeGenerated > ago(24h)

 Output:
 - MFA activity summary
 - Sign-in success and failure summary with distinct IP addresses
 - User risk level and risk state summary
 - Suspicious process execution summary
 - Correlated identity-to-endpoint insights
 - Concise triage summary report with investigation-ready findings

Refer to existing published agents in the Security Store such as Silverfort Identity Threat Triage Agent for formatting examples.

Measure SCU consumption

The plan description in Partner Center is required to include an SCU consumption estimate. Before creating your offer in Partner Center, run your agent 3–5 times under typical scenarios and record the SCU usage shown after each run in Security Copilot. Take the average and round up. You will add this to the plan description, for example, "This agent typically consumes 1.0 SCU per analysis run."

Create and configure your offer in Partner Center

Create a new SaaS offer in Microsoft Partner Center and configure the offer metadata, listing, and technical configuration.

Access Microsoft Partner Center

To access Microsoft Partner Center, follow these steps:

Go to Microsoft Partner Center
Sign in with your credentials
Navigate to Marketplace offers

Create new SaaS offer or clone existing offer

To create a new offer, follow these steps:

Select New offer
Select Software as a Service (SaaS) as the offer type
Select Start with a blank offer or Clone an existing offer

Tip

If you have an existing SaaS offer in Partner Center, clone it to reduce setup time. Cloning automatically carries forward your logo, legal documentation, privacy policy links, and other common metadata. You only need to update the Security Copilot agent specific content.
Enter your offer details, using these values as a reference:
- Offer ID: identity-drift-agent (lowercase, hyphens)
- Alias: IdentityDrift Investigation Agent
Select Create

Configure offer setup

Set up your offer in Partner Center by filling in the required fields. For an overview see Publish a Security Copilot Agent in Security Store.

On the Offer setup page enter the following values:

Would you like to sell through Microsoft?, Select Yes
Would you like to use Microsoft license management?, Select No
Customer leads: Optional. Connect your CRM if desired
Microsoft integrations: Check "My offer integrates with Microsoft Security services"
Select Save draft

Important

Enable Microsoft Security Services Integration
Checking "My offer integrates with Microsoft Security services" is required. Without enabling this checkbox, the "Microsoft Security services" option won't appear in the left navigation menu of your offer. This is the section where you upload your agent package .zip file.

Fill in the metadata properties

To configure metadata properties, follow these steps:

Go to Properties
Under Categories: Select Security or Compliance (primary category)
Industries: Leave blank
Legal contract: Choose Standard Contract or provide your own
Select Save draft

Configure the offer listing

To fill in the offer listing, follow these steps:

Go to Offer listing in the left menu
Fill in the required fields as follows:

Search results summary (single line): "Investigation agent that automates security incident investigation and response"

Description: prepared earlier in a structured format outlining agent tasks, inputs, and outputs.

Add images:
- Logo (216 x 216 px)
- Screenshots (1280 x 720 px)
Upload the user guide under Product information documents
Select Save draft

Best practices for offer setup

Important

Agent names must not contain Microsoft product names. The third-party agent name in Partner Center must not contain any Microsoft product names, including "Security Copilot", "Microsoft Sentinel", "Microsoft Defender", "Entra", etc.

Incorrect: "Contoso Security Copilot Investigation Agent" - Contains "Security Copilot"
Correct: "Contoso Identity Threat Triage Agent"

Check all locations including the offer name, the plan name, and the description text.

Offer listing: links and user guide.

Tip

The following two fields in the offer listing that are commonly missed and cause review failures:

Marketing/Product page link: Under Offer listing > Supplemental product information for customers > Product information links > Links, add the URL to your product marketing page or documentation page.
User guide document: Upload user guide PDF under Offer listing > Supplemental product information for customers > Product information documents. This document must include details for users to learn more about the agent and find instructions to install or use it.

For reference, see the Silverfort Identity Threat Triage Agent User Guide

Screenshots must show full agent execution and results. Screenshots are validated by the review team. Screenshots that show only configuration screens, setup pages, or UI without a running agent are rejected. At least one screenshot must:
- Show the agent actively running and returned results.
- Show the integrated Microsoft product (e.g., Microsoft Sentinel) visibly listed under the Plugins section in the agent view
- Screenshot resolution must be 1280×720 px. Use https://imageresizer.com to resize if needed.
To get Microsoft Sentinel to appear under Plugins, Add MCP.Sentinel to RequiredSkillsets in your AgentManifest.yaml.

Add Microsoft Security services metadata

To configure the Microsoft Security services section, follow these steps:

Go to Microsoft Security services in left menu

Configure metadata as follows:

Field	Value
Integrated Security services	Microsoft Security Copilot Microsoft Sentinel (as applicable)
Product prerequisites	Microsoft Security Copilot Microsoft Sentinel, Microsoft Defender, Microsoft Entra (as applicable)
Solution type	Deployable solution
License management	Choose based on your model

Check "Security Copilot agent"
Upload Solution Package:
- Select Upload .zip package
- Select your agent-package.zip file
Select Save draft

Tip

The Integrated Microsoft Security Products selection must accurately reflect what your agent actually integrates with in the description and in practice. Common mismatches that are hard fails are as follows:

Agent description mentions Sentinel data lake queries but only "Security Copilot" is selected → Add "Microsoft Sentinel"
Agent description mentions Defender alerts but "Microsoft Sentinel" is selected → Add "Microsoft Defender"

The selection drives which filter the agent appears under in the Security Store (for example, users filtering by "Microsoft Sentinel" will only see agents that have Microsoft Sentinel selected here).

Add preview audience

To add preview audience members, follow these steps:

Go to Preview audience in left menu
Enter Microsoft Entra IDs of internal users who will test:
- Team members
- QA testers
Select Save draft

Access preview

Invited preview audience members can access the offer listing as follows:

Users added to preview audience can access:
- Landing page
- Offer listing
- Full deployment flow
Share preview URL with test audience
Gather feedback on:
- Listing accuracy
- Deployment steps
- Documentation

Reference: How to preview and test your offer listing for Security Store

Technical configuration

To configure technical settings, follow these steps:

Go to Technical configuration in left menu

Fill in required fields:

Field	Value
Landing page URL	`https://securitystore.microsoft.com/mysolutions`
Connection webhook	Your webhook URL for order/subscription notifications
Microsoft Entra tenant ID	Your tenant ID
Microsoft Entra app ID	Your app ID

Select Save draft

Technical Configuration is mandatory and blocks submission. Partner Center will block the "Review and Publish" button if the technical configuration is not fully filled in, even if your organization does not plan to use telemetry or license management webhooks. This is a required section for all SaaS offers on the Microsoft commercial marketplace. If you are not ready to implement a webhook:

Use dummy/placeholder values for the landing page URL, webhook, tenant ID, and app ID
These values can be updated later by modifying and republishing the same offer

For webhook implementation guidance, see Implementing a webhook on the SaaS service - Marketplace publisher.

Create plan and pricing

Free agents

To configure a free agent plan, follow these steps:

Go to Plan overview in left menu
Select Create new plan
Enter the Plan name: Identity Drift Investigation Agent plan. Don't include Microsoft product names in plan name.
Enter the Plan description: Free tier offering of our agent. Include the SCU consumption estimate.
Select Edit Markets
Select regions where you want to offer (recommend Select all)
Set Pricing model: Flat rate
Set Contract duration, Billing Frequency, Price per charge: to $0 USD
Set Plan visibility to Public (or Private for specific customers)
Select Save draft

Paid agents

In addition to the steps in the Free agents section, include the following steps to configure a paid agent plan:

Pricing model: Choose between:
- Flat rate - Fixed monthly/annual fee
- Per user - Price per licensed user
Contract duration: 1 month, 1 year, 3 years, etc.
Billing frequency: Monthly or annual
Price per charge: Set your pricing
Optional: Check Free trial for 1-month trial period
Plan visibility: Set to Public (or Private for specific customers)
Select Save draft

Best practices for plan and pricing

Plan description must include SCU consumption estimate. The plan description is required to include an estimate of SCU consumption. Add a clear statement such as: "This agent typically consumes 1.0 SCU per analysis run."

To learn more about how to estimate SCU usage, see Manage usage - Microsoft Security Copilot.

The following is a sample full plan description:

The Contoso Identity Threat Triage Agent is available at no cost. This agent typically consumes 1.0 SCU per analysis run. SCU consumption may vary depending on the volume of data in your Microsoft Sentinel workspace and the complexity of the investigation.

Add supplemental content

To add supplemental content, follow these steps:

Select SaaS Scenarios - SaaS solution is not hosted in Azure.
In the text box, enter the following note: Offer listing is for Security Copilot Agent in Microsoft Security Store.
Upload Product documentation and select category Architecture diagram

Publish your agent

Final review checklist

Before publishing, verify:

All required fields completed
Offer listing is grammatically correct and accurate
Agent name does NOT contain Microsoft product names ("Security Copilot", "Microsoft Sentinel", etc.)
Marketing/product page link added under Offer Listing → Links
User guide document uploaded under Product information documents
Logo (216×216 px) and screenshots (1280×720 px)
At least one screenshot shows full agent execution with integrated product visible under Plugins
In AgentManifest.yaml file - in AgentDefinition section product and publisher fields reflect actual ISV name (not "Custom")
Plan description includes SCU consumption estimate
Technical configuration has all required values
ZIP package is valid and up to date
Terms and privacy policy links work

Submit for publication

To submit your offer for publication, follow these steps:

Select Review and publish (top right of Partner Center)
Review all sections
If all validation passes, select Publish
Your offer enters automated review process
After the automated review, select Go Live to move the request to Security Store team for review and certification

Monitor publication status

To monitor the status of your publication, follow these steps:

Return to Partner Center → Marketplace Offers
Find your offer
Review status indicators:
- In review - Being validated by Security Store team
- Changes required - Review feedback received, corrections needed
- Published - Live in Security Store

Verify live listing

Once published, verify your listing at Security Store.

Check:

Agent description displays correctly
Screenshots are visible
Links are working

Troubleshooting

Package zip file rejected or fails validation

Symptom	Solution
Invalid package structure	Verify `PackageManifest.yaml` is in root directory with correct formatting
Hidden files included in ZIP	On Mac, use: `zip -r package.zip . -x ".*" -x "__MACOSX"`
Agent name contains Microsoft product name	Remove `"Security Copilot"`, `"Microsoft Sentinel"`, `"Microsoft Defender"`, `"Entra"`, or any Microsoft product name from the agent name field in Partner Center. - Check all locations: the Offer name, Plan name, and description text body - Resubmit after renaming
Screenshots rejected — "do not depict functionality of the agent"	Screenshots must show a full agent execution run, not just configuration or setup pages - At minimum one screenshot must show: (1) agent actively running/returning results, (2) integrated Microsoft product (for example, Microsoft Sentinel) visible under Plugins - Add `MCP.Sentinel` to `RequiredSkillsets` in `AgentManifest.yaml` if Sentinel is not appearing in agent view

Your agent is now published to the Microsoft Security Store and available for customers. Monitor agent adoption and usage metrics, and use customer feedback to refine agent instructions and expand capabilities.

For more detailed information, see:

Feedback

Was this page helpful?

Last updated on 2026-07-20