AI Agent Workers
Status: Documentation Last Updated: 2025-12-27Overview
AI Agent Workers are the execution layer of Agency services. They perform automated tasks on behalf of clients using the same AI infrastructure that powers the Trending Engine platform.Agent Types
Content Agent
Creates and optimizes content for client channels.| Capability | Description |
|---|---|
| Blog posts | SEO-optimized articles |
| Social media | Platform-specific content |
| Email campaigns | Newsletter and drip sequences |
| Product descriptions | E-commerce copy |
Research Agent
Gathers intelligence and competitive insights.| Capability | Description |
|---|---|
| Market research | Industry trends and sizing |
| Competitor analysis | Positioning and pricing |
| Keyword research | SEO opportunity discovery |
| Audience insights | Demographics and behavior |
Automation Agent
Builds and maintains automated workflows.| Capability | Description |
|---|---|
| Workflow design | n8n/Make/Zapier builds |
| Integration setup | API connections |
| Data pipelines | ETL processes |
| Monitoring | Alert configuration |
Analysis Agent
Processes data and generates insights.| Capability | Description |
|---|---|
| Performance reports | KPI dashboards |
| Trend analysis | Historical patterns |
| Forecasting | Predictive models |
| Attribution | ROI tracking |
Execution Model
Task Delegation
Linear Integration
Agent work is tracked via Linear with delegation labels:| Label | Meaning |
|---|---|
delegate:cursor | Cursor agent executes |
delegate:claude-code | Claude Code executes |
delegate:gpt | ChatGPT executes (research) |
delegate:blocked | Needs human decision |
Agent Capabilities
Supported Tools
| Tool | Purpose | Agent Access |
|---|---|---|
| MCP Server | Database queries, tool execution | All agents |
| Web Search | Research and fact-checking | Research, Content |
| File System | Read/write project files | Cursor, Claude Code |
| Linear API | Issue management | All agents |
| Supabase | Data operations | Automation, Analysis |
Quality Controls
Each agent output goes through validation:Client Configuration
Agent Permissions
Each client has configured agent access:Usage Tracking
Pricing Model
Cost Structure
| Component | Cost | Passed to Client |
|---|---|---|
| LLM tokens | Variable | At cost + 20% |
| Tool usage | Per call | Included in package |
| Human review | Per hour | $75/hour |
Package Allocation
| Package | Agent Hours/Month | Included Reviews |
|---|---|---|
| Starter | 10 | 2 |
| Growth | 25 | 5 |
| Scale | 50 | 10 |
| Enterprise | Unlimited | Unlimited |
Security & Compliance
Data Isolation
- Each client’s data is isolated by
tenant_id - Agents only access client-authorized resources
- Credentials stored encrypted in Supabase Vault
Audit Trail
All agent actions are logged:Human-in-the-Loop
Critical actions require human approval:| Action Type | Auto-Execute | Requires Approval |
|---|---|---|
| Content draft | ✓ | |
| Research summary | ✓ | |
| Publish content | ✓ | |
| Data deletion | ✓ | |
| API credential use | ✓ | |
| External API calls | ✓ (first time) |
Performance Metrics
Agent Efficiency
Quality Metrics
| Metric | Target | Measurement |
|---|---|---|
| First-pass acceptance | > 80% | Outputs approved without revision |
| Revision rounds | < 2 | Average revisions per deliverable |
| Client satisfaction | > 4.5/5 | Post-delivery survey |
Related Documentation
| Document | Purpose |
|---|---|
| OVERVIEW.md | Agency architecture |
| automation-packages.md | Package definitions |
| deliverables.md | Output management |
| reporting.md | Client reporting |