Agent Types

Overview

Logrite agents are categorized by their origin, execution model, and permission level.

Type 1: System Agents

Pre-built agents that ship with the platform. They are read-only (users cannot edit or delete them) and are scoped to the platform level.

Agent	Schedule	Tools	Purpose
Anomaly Monitor	Every 15 min	opensearch_query	Checks OpenSearch anomaly detection results
Error Rate Watchdog	Every 5 min	semantic_search, opensearch_query	Monitors error rates across all indexes
Compliance Auditor	Daily 6 AM	get_scan_results, analyze_logs	Summarizes recent compliance scan findings
Health Reporter	Hourly	opensearch_query	Generates service health digest

System agents post results to the shared Discord channel but cannot create scan records or modify data.

Type 2: User-Created Agents

Agents created by workspace users via the AgentBuilder UI. These agents are:

• Owned by the workspace
• Scoped to workspace data only
• Limited to toolsAllowed as set during creation
• Subject to workspace rate limits

Sub-types by creation method:

Sub-type	How Created	Config Source
NLP-created	Natural language description	AI Service parses intent
Template-based	Start from a template	Pre-defined config, user customizes
Manual	Form-based creation	User fills fields directly

Type 3: Scheduled Agents

Any agent (system or user-created) with a cron schedule. The Agent Runner manages their execution:

"schedule": "0 * * * *"     # Every hour
"schedule": "0 8 * * 1-5"   # Weekdays at 8am
"schedule": "*/15 * * * *"  # Every 15 minutes
"schedule": "manual"         # No automatic execution

Type 4: Event-Triggered Agents (Phase 4)

Agents that fire in response to events rather than time:

Trigger	Example
Anomaly detected	Payment latency spike > 0.6 grade
Error rate threshold	>10 errors in 5 minutes
Scan complete	New compliance scan finished
Log pattern match	Specific string detected in logs

Agent States

State	Description
draft	Created but not yet enabled
active	Running on schedule
paused	Temporarily disabled
error	Last run failed, requires attention

Permissions by Agent Type

Capability	System	User-Created	Admin
Read own workspace logs	✅	✅	✅
Read other workspace logs	❌	❌	✅
Create scan records	❌	❌	✅
Trigger deployments	❌	❌	❌ (never)
Send Discord alerts	✅	✅ (own channel)	✅
Modify agent configs	❌	Own agents only	✅