m4Mindset docs
llms.txt

Mindset v3 / Build & Optimize / AMS SDK

View as Markdown

End-to-end Agent Deployment Flow

From tenant creation to end-user access

Overview: Two-Part Process

Part 1: Enable Creation

Tenants create agents and contexts using Agent Builder SDK components

Part 2: Deploy to Users

You deploy agents to end-users via Agent Sessions API

Part 1: Enable Tenant Creation

  1. Configure in Agent Builder SDK Playground

    • Access Developer Tools → Agent Builder SDK Playground
    • Preview components (Agent Builder, Context Builder)
    • Customize theme, colors, fonts
    • Select which configuration tabs to display
    • Generate integration code

  2. Embed Components in Your Platform

    For complete UI with list view:

    • <mindset-agents-manager>
    • <mindset-contexts-manager>

    For custom list view:

    • <mindset-agent-configuration>
    • <mindset-context-configuration>

    Flexible embedding:

    • Embed single components per page or combine multiple components on the same page

  3. Generate Authentication Token for Tenant Admin

    Before the tenant admin can use Agent Builder SDK components:

    • Generate an authToken for the tenant admin user
    • CRITICAL: When creating the authToken, explicitly designate the user as a tenant admin by passing the externalTenantId
    • Provide this authToken to the Mindset SDK that runs in the tenant admin's page

  4. Tenant Creates Agent

    Tenant Admin (externalTenantId: "tenant-acme-123") → Creates Agent via Agent Builder or Agent Configuration

  5. Tenant Creates Contexts & Uploads Files (Optional)

    • Creates Contexts via Context Builder or Context Configuration
    • Uploads Files (PDFs, docs, training materials, etc.)

Part 2: Deploy to End-Users

  1. Your Backend Determines Permissions

    Based on your platform's permission system, determine for each end-user:

    • Which agent(s) should they access?
    • Which contexts/knowledge should they see?
    • Which MCP tools should be available?

  2. Create Agent Session via API

    Required:

    • agentUid (tenant's agent)
    • externalUserId

    Choose ONE:

    • contextUids (up to 30)
    • OR 1 RAG MCP

    Optional:

    • mcpserverUids (tools, up to 5 total)

    Recommended:

    • tags (include externalTenantId)

  3. Receive agentSessionUid

    API returns unique session identifier Example: agentSessionUid: "agent-123::session-789"

  4. Generate Authentication Token for End-User

    Before the end-user can access the agent:

    • Generate an authToken for the end-user
    • CRITICAL: The user's externalUserId in the authToken must be an exact match to the externalUserId used when creating the agent session for them
    • Provide this authToken to the Agent SDK component

  5. Embed Agent Chat SDK for End-User

    In the end-user's UI, embed:

    html
            <mindset-agent agentUid="agent-123::session-789"></mindset-agent>

    ✅ Deployment Complete!

Key Concepts

Authentication Requirements

For Tenant Admins (Agent Builder SDK):

  • Generate authToken with externalTenantId specified
  • This designates them as tenant admin with access to Agent Builder SDK components

For End-Users (Agent SDK):

  • Generate authToken with externalUserId
  • externalUserId must exactly match the one used in the agent session creation
Tenant Isolation
  • externalTenantId filters what tenants see in Agent Builder SDK components
  • Your backend enforces isolation in agent sessions
  • The same externalTenantId in the Agent Builder SDK components means users see each other's agents/contexts in the list views
  • Different externalTenantId values mean complete isolation in the list views
Flexible Context Mixing
  • Agent sessions can combine app-level and tenant-level contexts
  • You control what each user sees through the Agent Sessions API
  • Up to 30 contexts can be provisioned per agent session
Per-User Permissions
  • Same agent can be deployed with different contexts and tools for different users
  • Based on your platform's permission system
  • Your backend determines what each end-user can access

Agent Session Composition Rules

Rule 1: Knowledge Sources (Mutually Exclusive)

Rule 2: MCP Servers (Maximum 5 Total)

RAG-type MCP: 0 or 1
Tools-type MCPs: 0 to 5
✅ Valid: 1 RAG MCP + 4 Tools MCPs = 5 total

Rule 3: Tags (Maximum 10)

Used for reporting, grouping, and filtering
Recommended: Include externalTenantId

Valid Configuration Examples

Configuration 1: Contexts + Tools
{
  "agentUid": "tenant-agent-123",
  "externalUserId": "user-456",
  "contextUids": ["context-1", "context-2", "context-3"],
  "mcpserverUids": ["gmail-tool", "slack-tool"],
  "tags": ["tenant-acme-123"]
}
Configuration 2: RAG MCP + Tools
{
  "agentUid": "tenant-agent-456",
  "externalUserId": "user-789",
  "mcpserverUids": ["company-rag-mcp", "salesforce-tool", "zendesk-tool"],
  "tags": ["tenant-techcorp-456"]
}
Configuration 3: Contexts Only
{
  "agentUid": "tenant-agent-789",
  "externalUserId": "user-012",
  "contextUids": ["context-1", "context-2", "...", "context-30"],
  "tags": ["tenant-widgets-789"]
}

Invalid Configurations