Skip to Content
AdministrationOverview

Administration

Platform administration for multi-tenant environments, user management, LLM provider configuration, and security settings.

Overview

M3 Forge provides comprehensive administration capabilities for managing:

  • Workspaces — Multi-tenant isolation with separate data and resources
  • Users and roles — RBAC with fine-grained permissions
  • LLM connections — Configure multiple AI providers and models
  • API keys — Programmatic access control and authentication
  • System settings — Theme, notifications, integrations

Administrators access these features via the Settings section in the navigation menu.

Administration dashboard showing system health, user count, and configuration status

Multi-Tenancy

M3 Forge supports multi-tenant deployments where each workspace provides:

  • Data isolation — Workflows, prompts, and configurations are workspace-scoped
  • Resource quotas — Set limits on storage, API calls, workflow executions
  • Team management — Independent user lists and role assignments per workspace
  • Billing separation — Track usage and costs per workspace

This enables:

  • SaaS deployments serving multiple organizations
  • Department isolation within large enterprises
  • Client separation for consulting firms

See Workspaces for details.

Role-Based Access Control

M3 Forge uses a role-based permission model with predefined and custom roles.

Built-in Roles

RolePermissionsUse Case
OwnerFull platform access including user managementWorkspace administrators
AdminAll features except user/workspace managementTechnical leads
EditorCreate and edit workflows, prompts, agentsEngineers and data scientists
ViewerRead-only access to all resourcesStakeholders and auditors
OperatorRun workflows, view monitoringOperations team members

Custom Roles

Create custom roles with specific permission combinations:

  • Workflow permissions — Create, read, update, delete, execute
  • Prompt permissions — Create, read, update, delete, test
  • Settings permissions — View, update system configuration
  • User permissions — Invite, manage roles, remove users

See Users and Roles for configuration.

LLM Provider Management

M3 Forge supports multiple LLM providers simultaneously, enabling:

  • Model diversity — Use different models for different tasks
  • Provider redundancy — Failover if one provider has an outage
  • Cost optimization — Route requests to most cost-effective provider
  • Geographic compliance — Use region-specific endpoints

Supported Providers

ProviderModelsFeatures
OpenAIGPT-4o, GPT-4 Turbo, GPT-3.5Function calling, vision, embeddings
AnthropicClaude 3.5 Sonnet, Claude 3 Opus, Claude 3 HaikuLong context, extended thinking
QwenQwen-2.5, Qwen-VLMultilingual, vision-language
Google Vertex AIGemini 1.5 Pro, Gemini 1.5 FlashMultimodal, long context
Azure OpenAIGPT-4o, GPT-4, GPT-3.5Enterprise features, compliance
BedrockClaude, Llama, TitanAWS-native deployment

Each provider connection can have multiple API keys for load balancing and failover.

See LLM Connections for setup.

API Key Management

Generate API keys for programmatic access to M3 Forge:

  • Workflow execution — Trigger runs via REST or tRPC
  • Data ingestion — Upload documents for processing
  • Monitoring integration — Query metrics and logs
  • CI/CD pipelines — Automated deployment of workflows

API keys support:

  • Scoped permissions — Limit keys to specific operations
  • Expiration — Set time-based validity
  • Rotation — Update keys without downtime
  • Audit logging — Track all API key usage

See API Keys for details.

Settings Organization

The Settings interface is organized into sections:

General

  • Workspace name — Display name for current workspace
  • Theme — Light, dark, or system preference
  • Language — Interface localization (English, Spanish, French, etc.)
  • Timezone — For timestamp display

Security

  • Session timeout — Automatic logout after inactivity
  • Password policy — Complexity requirements
  • Two-factor authentication — Enforce 2FA for all users
  • IP allowlist — Restrict access to specific networks

Integrations

  • Webhooks — Outbound notifications for events
  • OAuth providers — SSO with Google, GitHub, Microsoft
  • SAML — Enterprise SSO integration
  • Audit log exports — Send logs to external SIEM

Notifications

  • Email alerts — Workflow failures, usage limits
  • Slack integration — Real-time notifications
  • Webhook notifications — Custom integrations

Billing

  • Usage dashboards — API calls, storage, compute
  • Quota management — Set limits per workspace
  • Invoice history — Download past invoices
  • Payment methods — Credit card, ACH, invoicing

Billing features are only available in multi-tenant SaaS deployments. Self-hosted installations do not include billing.

Admin Dashboard

The Admin Dashboard provides at-a-glance monitoring:

System Health

  • Service status — API server, database, storage, queue
  • Response times — 95th percentile latency
  • Error rates — Failed requests per minute
  • Uptime — Service availability percentage

Usage Metrics

  • Active users — Unique users in last 24 hours
  • Workflow executions — Runs per day
  • LLM API calls — Requests to external providers
  • Storage utilization — Documents, logs, artifacts

Resource Consumption

  • CPU usage — Across all services
  • Memory usage — Database, API server, workers
  • Disk I/O — Read/write throughput
  • Network bandwidth — Ingress/egress

Recent Activity

  • New users — Recent registrations
  • Failed logins — Potential security issues
  • Workflow errors — Recent failures
  • API key usage — Top consumers

Getting Started

Best Practices

Workspace Structure

For enterprise deployments:

  • One workspace per department — Finance, HR, Legal each get isolated environments
  • Shared workspace for common resources — Company-wide prompt library, shared workflows
  • Development and production workspaces — Separate environments for testing vs production

Role Assignment

Follow the principle of least privilege:

  • Start users with Viewer role and grant additional permissions as needed
  • Use Editor role for day-to-day work, not Admin
  • Reserve Owner role for 1-2 people per workspace
  • Create custom roles for specialized needs (e.g., “Prompt Engineer” with full prompt access but no workflow editing)

LLM Connection Management

  • Use provider-specific API keys — Don’t share keys across providers
  • Set rate limits — Prevent accidental spend on expensive models
  • Monitor usage — Set up alerts for unusual consumption
  • Rotate keys quarterly — Regular key rotation for security

API Key Hygiene

  • Set expiration dates — Force periodic rotation
  • Scope narrowly — Each key should have minimal necessary permissions
  • Name descriptively — Include purpose and owner (e.g., “CI-Pipeline-Deploy-Key”)
  • Revoke unused keys — Clean up old or forgotten keys

Security Considerations

Authentication

M3 Forge supports:

  • Email/password with strong password policies
  • OAuth 2.0 — Google, GitHub, Microsoft SSO
  • SAML — Enterprise identity providers
  • Two-factor authentication — TOTP or SMS

Authorization

  • All actions are permission-checked against user roles
  • Workspace isolation enforced at database query level
  • API keys inherit permissions from creating user
  • Admin actions are logged for audit

Data Protection

  • Encryption at rest — Database and storage encrypted
  • Encryption in transit — TLS for all API communication
  • Secret management — API keys stored hashed
  • PII handling — Configurable data retention and anonymization

Compliance

M3 Forge provides features for:

  • SOC 2 — Audit logging, access controls
  • GDPR — Data export, right to deletion
  • HIPAA — PHI handling with BAA support
  • ISO 27001 — Security policies and procedures

See contract/security.md in the Marie ecosystem for complete security standards.

Troubleshooting

Cannot Access Settings

Cause: User lacks admin permissions.

Solution: Workspace owner must grant admin or owner role.

LLM Connection Failing

Cause: Invalid API key or network connectivity issue.

Solution: Test connection in LLM Connections settings. Check API key validity with provider.

Users Not Receiving Invites

Cause: Email delivery issues or spam filtering.

Solution: Check email settings under Integrations. Ask users to check spam folder. Use OAuth instead of email invites.

API Key Not Working

Cause: Key expired, insufficient permissions, or incorrect scope.

Solution: Regenerate key with correct permissions and expiration. Check audit logs for failed authentication attempts.

Next Steps

Last updated on