Skip to content

Latest commit

 

History

History
473 lines (353 loc) · 15.5 KB

AGENT_PORTALS.md

File metadata and controls

473 lines (353 loc) · 15.5 KB

Agent Portals

Each of the 12 AI agents in the Rinse Repeat Labs orchestrator has a dedicated portal in the web dashboard. These portals allow agents to manage their workspace and request new features, just like employees in a real company.

Overview

Agent portals serve two key purposes:

  1. Workspace Management — Each agent has a customized view of their responsibilities, metrics, and relevant data
  2. Feature Requests — Agents can request new tools, data, or improvements to their section of the dashboard

This creates a self-improving system where agents can identify their own needs and the Architect (human overseer) can prioritize and implement the most valuable changes.

Accessing Agent Portals

Web Dashboard

Start the Flask web application:

cd webapp
python app.py

Navigate to:

  • All Portals: http://localhost:5000/agents
  • Individual Portal: http://localhost:5000/agents/<agent_id>

Agent IDs: ceo, cfo, cito, sales, legal, dev_lead, design_lead, qa_lead, pm, customer_success, marketing, support

CLI Access

# List all feature requests
python orchestrator.py requests list

# View pending requests
python orchestrator.py requests pending

# Show a specific request
python orchestrator.py requests show <request_id>

Portal Components

Agent Profile

Each portal displays:

  • Role & Name — Agent's title and full name
  • Description — What the agent does
  • Team — Executive, Technical, or Operations
  • Responsibilities — Key duties
  • Metrics — Performance indicators they track

Feature Requests

Agents can view and manage their submitted requests:

  • Total Requests — All requests submitted
  • Pending — Awaiting review
  • Approved — Ready for implementation
  • Implemented — Completed and deployed

Recent Activity

Timeline of recent actions and request status changes.

Portal Tabs

Each agent portal has multiple tabs for different functionality:

Overview Tab

The main tab showing:

  • Agent profile and description
  • Key responsibilities and metrics
  • Custom instructions (if configured)
  • Recent activity timeline
  • Quick actions and navigation

Settings Tab

Customize how the agent appears and behaves:

Setting Description
Display Name Custom name shown in dashboard
Role Title Short title shown in badges
Description What this agent does
Responsibilities Key duties (one per line)
Key Metrics Performance indicators (one per line)
Custom Instructions Extra context for AI meetings
Reports To Who this agent reports to
Direct Reports Agents who report to this one
Collaborates With Key collaboration relationships

Settings are stored in data/agent_customizations.json and affect both the web interface and CLI meetings.

Feature Requests Tab

View and manage the agent's submitted requests (see Feature Request System).

How It Works Tab

Documentation explaining how this agent functions within the orchestrator system.

Web-Based Agent Chat

Each agent has a chat interface for real-time conversations through the web dashboard.

Accessing Chat

  1. Navigate to an agent's portal: /agents/<agent_id>
  2. Click "Start Chat" in Quick Actions
  3. Or go directly to: /agents/<agent_id>/chat

Chat Features

  • Real-time messaging — Send messages and receive AI responses
  • Conversation history — Chat sessions are persisted and can be resumed
  • Agent context — The AI has access to the agent's role, responsibilities, and custom instructions
  • Conversation starters — Quick prompts to begin discussions

CEO Chat: Special Powers

The CEO agent has enhanced capabilities through the chat interface:

Capability Description
Update Agent Settings Modify any agent's display name, role, instructions, etc.
Update Company Settings Change company name, tagline, industry
Create Feature Requests Submit requests on behalf of agents
Approve/Reject Requests Process pending feature requests
Update Idea Status Move ideas through the pipeline
Broadcast Instructions Add instructions to multiple agents at once
Update Reporting Structure Change who reports to whom

When the CEO proposes an action, it appears as a confirmation panel. You must Confirm or Dismiss before the action executes.

CEO Action Format

The CEO can include actions in responses using this format:

[ACTION: action_type]
{"parameter": "value"}
[/ACTION]

Example CEO interaction:

You: "Update the CFO to focus on bootstrapped startups"

CEO: "I'll update the CFO's custom instructions to emphasize
bootstrapped startup financial strategies.

[ACTION: update_agent_settings]
{"agent_id": "cfo", "updates": {"custom_instructions": "Focus on bootstrapped startups with limited runway. Emphasize capital efficiency and sustainable growth over aggressive spending."}}
[/ACTION]"

→ Confirmation panel appears
→ Click "Confirm" to execute
→ CFO's settings are updated

Available CEO Actions

Action Parameters Description
update_agent_settings agent_id, updates Update agent configuration
update_company_settings updates Change company name/tagline/industry
create_feature_request agent_id, title, description, priority, request_type Create a feature request
approve_feature_request request_id, notes Approve a pending request
reject_feature_request request_id, reason Reject a pending request
update_idea_status idea_id, new_status, notes Change an idea's status
broadcast_to_agents agent_ids, instruction, append Add instructions to multiple agents
update_reporting_structure agent_id, reports_to, direct_reports Change reporting hierarchy

Chat Data Storage

Chat sessions are stored in data/agent_chats.json:

{
  "id": "uuid",
  "agent_id": "ceo",
  "topic": "Strategic planning",
  "messages": [
    {"role": "user", "content": "...", "timestamp": "..."},
    {"role": "assistant", "content": "...", "timestamp": "..."}
  ],
  "started_at": "2026-01-20T10:00:00",
  "ended_at": null,
  "is_active": true
}

Web-Based Group Meetings

In addition to 1:1 chat with agents, you can hold group meetings with multiple agents through the web dashboard.

Accessing Group Meetings

  1. Navigate to Agent Portals > Start Group Meeting
  2. Or go directly to: /group-meetings/new
  3. Or from the agents list page, click Start Meeting

Meeting Presets

Choose from preset meeting configurations or create custom combinations:

Preset Agents Best For
Executive Meeting CEO, CFO, CITO, Sales, Legal Strategic business decisions
Technical Meeting CITO, Dev Lead, Design Lead, QA Lead Architecture, technical quality
Product Meeting PM, Dev Lead, Design Lead, QA Lead Sprint planning, product features
Operations Meeting PM, Customer Success, Marketing, Support Day-to-day operations
All-Hands Meeting All 12 agents Company-wide alignment
Idea Review CEO, CITO, CFO, Dev Lead, Design Lead, Marketing Evaluating new ideas
Custom Meeting Your selection Any agent combination

Meeting Features

  • Multi-agent responses — All participants respond to your messages in turn
  • Role-specific perspectives — Each agent contributes their unique viewpoint
  • Quick prompts — Built-in prompts for summarizing, identifying risks, next steps
  • Conversation history — Meetings are saved and can be reviewed later
  • Active/Ended status — End meetings when done, start new ones anytime

Group Meeting Data Storage

Meetings are stored in data/meetings.json:

{
  "id": "uuid",
  "meeting_type": "exec",
  "meeting_name": "Executive Meeting",
  "topic": "Q1 Planning",
  "agent_ids": ["ceo", "cfo", "cito", "sales", "legal"],
  "messages": [
    {"role": "user", "content": "...", "timestamp": "..."},
    {"role": "ceo", "content": "...", "agent_name": "Alex Chen", "timestamp": "..."}
  ],
  "started_at": "2026-01-20T10:00:00",
  "ended_at": null,
  "is_active": true
}

Feature Request System

Request Types

Type Description Examples
Feature New functionality Dashboard widgets, new tools
Enhancement Improvements to existing features Better filtering, faster loading
Data/Report New metrics, dashboards, reports KPI tracking, trend analysis
Integration External system connections API integrations, third-party tools
UI/UX Interface improvements Better layouts, clearer navigation
Automation Automated workflows Alerts, scheduled reports, reminders

Priority Levels

Priority Description Use When
Critical Blocking current work Can't perform essential duties
High Significantly impacts effectiveness Major productivity improvement
Medium Would improve workflow Moderate benefit (default)
Low Nice to have Minor convenience

Request Workflow

┌──────────┐     ┌─────────────┐     ┌──────────┐     ┌─────────────┐     ┌─────────────┐
│ Submit   │────▶│ Under Review│────▶│ Approved │────▶│ In Progress │────▶│ Implemented │
└──────────┘     └─────────────┘     └──────────┘     └─────────────┘     └─────────────┘
                        │
                        │
                        ▼
                 ┌──────────┐
                 │ Rejected │
                 └──────────┘
                        │
                        ▼
                 ┌──────────┐
                 │ Deferred │
                 └──────────┘
  1. Submitted — Request created by agent
  2. Under Review — Architect is evaluating
  3. Approved — Request accepted, awaiting implementation
  4. Rejected — Request declined (with reason)
  5. Deferred — Postponed for later consideration
  6. In Progress — Currently being implemented
  7. Implemented — Complete and deployed

Agent Voting

Other agents can vote on requests:

  • Support — Indicates the feature would help them too
  • Oppose — Indicates concerns about the request

Votes help the Architect understand cross-team impact and prioritize effectively.

CLI Commands

Listing Requests

# All requests
python orchestrator.py requests list

# Filter by agent
python orchestrator.py requests list --agent ceo

# Filter by status
python orchestrator.py requests list --status approved

# Limit results
python orchestrator.py requests list --limit 10

Viewing Requests

# Show pending requests (prioritized)
python orchestrator.py requests pending

# Show specific request details
python orchestrator.py requests show <request_id>

Creating Requests

# Create a feature request
python orchestrator.py requests create \
  --agent ceo \
  --title "Revenue Dashboard" \
  --description "Real-time dashboard showing revenue by project and client" \
  --type feature \
  --priority high \
  --justification "Need visibility into financial performance for strategic decisions" \
  --area "Dashboard"

Managing Requests (Architect)

# Approve a request
python orchestrator.py requests approve <request_id> \
  --note "Great idea, adding to next sprint"

# Reject a request
python orchestrator.py requests reject <request_id> \
  --reason "Duplicates existing functionality"

# Update status
python orchestrator.py requests status <request_id> in_progress \
  --note "Development started"

# Cast a vote (as another agent)
python orchestrator.py requests vote <request_id> cfo --type support

Best Practices

For Agents

When submitting requests:

  1. Be Specific — Clearly describe what you need
  2. Explain the Why — Include business justification
  3. Consider Impact — Think about how it affects other agents
  4. Set Realistic Priority — Reserve "critical" for true blockers
  5. Suggest Solutions — Include ideas for implementation

Example of a good request:

Title: Client Health Score Dashboard

Description: A dashboard showing client health scores based on:
- Project delivery status
- Payment history
- Support ticket trends
- NPS/satisfaction scores

Justification: Currently I have to manually check multiple systems
to assess client health. An aggregated view would save 2-3 hours
weekly and help identify at-risk clients before issues escalate.

Type: Data/Report
Priority: High
Affected Area: CustomerSuccess Dashboard

For the Architect

When reviewing requests:

  1. Consider Cross-Team Value — Look at votes from other agents
  2. Batch Related Requests — Implement related features together
  3. Provide Feedback — Always explain approval/rejection decisions
  4. Track Patterns — Repeated requests indicate real needs
  5. Balance Teams — Ensure all teams get attention

Agent-Specific Portals

Executive Team

Agent Portal Focus
CEO Strategic metrics, company health, key relationships
CFO Financial dashboards, invoicing, cash flow
CITO Technical evaluation, architecture decisions, team metrics
Sales Pipeline management, lead tracking, deal velocity
Legal Contract tracking, compliance status, NDA management

Technical Team

Agent Portal Focus
DevLead Code quality, sprint velocity, technical debt
DesignLead Design assets, user research, accessibility
QALead Test coverage, bug metrics, release readiness

Operations Team

Agent Portal Focus
PM Project status, resource allocation, delivery metrics
CustomerSuccess Client health, NPS, retention metrics
Marketing Growth metrics, ASO rankings, campaign performance
Support Ticket volume, response times, knowledge base

Data Storage

Feature requests are stored in data/agent_requests.json with the following structure:

{
  "id": "uuid",
  "agent_id": "ceo",
  "title": "Request Title",
  "description": "Detailed description",
  "request_type": "feature",
  "priority": "medium",
  "status": "submitted",
  "justification": "Business justification",
  "affected_area": "Dashboard section",
  "votes": {
    "support": ["cfo", "sales"],
    "oppose": []
  },
  "review_notes": [],
  "created_at": "2026-01-19T12:00:00",
  "reviewed_at": null,
  "reviewed_by": null
}

Integration with Meetings

Agents are aware of their portal capabilities during meetings. When an agent identifies a need for new functionality, they can:

  1. Mention it in the meeting for discussion
  2. Submit a formal request through their portal
  3. Get feedback from other agents (votes)
  4. Track implementation progress

This creates a feedback loop where meeting discussions can drive portal improvements, and portal analytics can inform meeting discussions.