Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.junis.ai/llms.txt

Use this file to discover all available pages before exploring further.

​
Welcome to the Admin Panel

The Junis Admin Panel is your central command center for managing AI agents, workflows, knowledge bases, and team settings. This no-code interface allows you to build sophisticated multi-agent systems without writing a single line of code. What You’ll Learn:
  • Navigate the Admin interface
  • Manage agents and workflows
  • Configure RAG knowledge bases
  • Connect MCP integrations
  • Monitor usage and performance

​
📍 Accessing the Admin Panel

1

Sign In

Navigate to https://junis.ai and sign in with your account
2

Select Organization

From the top navigation bar, ensure you’ve selected the correct organization (if you’re a member of multiple organizations)
3

Access Admin

Click “Admin” in the left sidebar navigation

​
Permission Levels

Different features are available based on your role:
FeatureMemberAdmin
View Dashboard✅✅
Manage API Keys✅✅
View Agents❌✅
Create/Edit Agents❌✅
Organization Settings❌✅
System Management❌❌
Role Hierarchy: Admin > Member. Higher roles inherit all permissions from lower roles.

​
📊 Dashboard Overview

The Dashboard is the first page you see when entering the Admin panel. It provides real-time insights into your organization’s AI usage.

​
Quick Stats

The top section displays four key metrics:

Active Users Today

Number of unique users in the last 24 hours

Active Chats Today

Total chat sessions created today

Live Now

Currently connected WebSocket sessions (real-time)

Avg Speed

Average response time in milliseconds (last 24 hours)

​
Real-Time Metrics

The metrics section shows live updates via WebSocket:
  • Active Sessions: Users currently chatting with agents
  • Recent Messages: Message rate per minute
  • Agent Performance: Response times by agent
  • Error Rate: System health monitoring
Metrics refresh automatically every few seconds. Look for the green indicator to confirm live connection status.

​
Recent Activity (Admin Only)

Admins and Super Admins can see:
  • Latest chat sessions
  • Which agents were used
  • Session duration and message count
  • User information

​
🤖 Agent Management

The heart of Junis - where you create, configure, and visualize your AI agent workflows.

​
Two Views: List and Orchestration

Table View - See all agents in a sortable, hierarchical listColumns:
  • Name: Agent identifier (Orchestrator shown first)
  • Type: LLM, Sequential, Parallel, or Loop (color-coded badges)
  • SubAgents: Visual chips showing connected agents with order numbers
  • Description: Agent purpose and function
  • Actions: Edit, Relationships, Delete buttons
Sorting: Agents are automatically sorted by hierarchy, starting with the Orchestrator, then its sub-agents in execution order.

​
Creating a New Agent

1

Click New Agent

From the Agent Management page, click the ”+ New Agent” button (top right)
2

Choose Method

Select how to create the agent:
  • Create New: Build from scratch
  • Use Global Agent: Import a public template from the Junis marketplace
3

Configure Basic Info

Basic Settings
Name: My New Agent
Agent Type: LLM_AGENT
Description: What this agent does
Status: Active ✅
4

Configure LLM Settings (for LLM Agents)

Model Configuration
Model: claude-haiku  # Or gpt, gemini, etc.
Temperature: 0.7
Max Output Tokens: 16000
Streaming: Enabled ✅
5

Write Instructions

System Prompt
You are a helpful assistant that...

Your responsibilities:
1. Answer user questions clearly
2. Use tools when appropriate
3. Be concise and professional

Important: Always verify information before responding.
6

Add Tools (Optional)

  • Regular Tools: Select Python functions (search_database, send_email, etc.)
  • RAG Tools: Select knowledge bases (rag_product_docs, rag_support_kb)
  • MCP Platforms: Select integrations (GitHub, Firecrawl)
7

Configure Advanced Settings

  • Output Key: Variable name for Sequential Agent data passing
  • Payment Settings: Enable paid usage (wallet address, price per use)
  • Web Search: Enable real-time web search (Gemini 2.0, Claude 3.5+)
  • Extended Thinking: Enable reasoning tokens (Claude 3.5+, o1 models)
8

Save

Click “Create” - your agent will be instantly available!
Auto-Refresh: The cache is automatically refreshed when you create or update agents. No manual reload needed!

​
Editing an Agent

  1. Click the Edit (pencil) icon next to the agent name
  2. Modify settings in the dialog
  3. Click “Save” to apply changes
Global Agents (marked with a globe icon) cannot be edited. These are public templates provided by Junis. To customize, create a copy instead.

​
Managing Agent Relationships

Agent Relationships define which sub-agents a parent agent can call.
1

Open Relationships Dialog

Click the GitBranch icon next to the agent name
2

View Current Relationships

See two sections:
  • Sub-Agents: Agents this one calls (with order index)
  • Parent Agents: Agents that call this one
3

Add a Sub-Agent

  1. Click ”+ Add Sub-Agent”
  2. Select an agent from the dropdown
  3. Set the order index (1 = first, 2 = second, etc.)
  4. Click “Add”
4

Reorder Sub-Agents

For Sequential Agents, order matters!
  • Click the up/down arrows to change execution order
  • Changes save immediately
5

Remove a Sub-Agent

Click the trash icon next to a sub-agent relationship
For Sequential Agents: Sub-agents execute in order (1, 2, 3…). Each agent passes its output to the next via the output_key variable.
For Parallel Agents: Order doesn’t matter. All sub-agents run simultaneously, and results are combined.

​
Visual Workflow Builder

The Orchestration tab provides an interactive graph view of your agent system. Node Colors:
  • Blue Gradient: LLM Agent (performs tasks)
  • Purple Gradient: Sequential Agent (runs agents in order)
  • Green Gradient: Parallel Agent (runs agents simultaneously)
  • Orange Gradient: Loop Agent (repeats until condition met)
Connection Lines:
  • Solid: Direct parent-child relationship
  • Dotted (Purple): Sequential execution flow (left to right)
  • Dotted (Green): Parallel execution flow (top to bottom)
Legend (top of the page):
  • Shows agent type meanings
  • Indicates execution flow patterns
Unconnected Agents Panel (right side):
  • Lists agents not connected to the Orchestrator
  • Displays active/inactive status
  • Tool Agents are automatically hidden
Pro Tip: Use the minimap (bottom right) to navigate large workflows. Click and drag the viewport box to move around.

​
Deleting an Agent

Deletion is permanent! Make sure you’re not deleting an agent that other agents depend on.
1

Check Relationships

Before deleting, open the Relationships dialog to see if other agents use this one
2

Remove Dependencies

If the agent has parent agents, remove those relationships first
3

Delete

Click the trash icon, confirm the deletion
Orchestrator Protection: You cannot delete the Orchestrator agent. It’s the root of your organization’s agent system.

​
📚 RAG Knowledge Base

RAG (Retrieval-Augmented Generation) allows your agents to query uploaded documents for accurate, context-aware responses.

​
How It Works

​
Uploading Documents

1

Go to RAG Management

Navigate to Team > RAG Knowledge in the sidebar
2

Prepare Your Files

Supported Formats:
  • ✅ PDF (.pdf) - Max 50MB
  • ✅ Microsoft Word (.docx) - Max 50MB
  • ✅ PowerPoint (.pptx) - Max 10MB
  • ✅ Plain Text (.txt) - Max 10MB
  • ✅ Markdown (.md, .markdown) - Max 10MB
  • ✅ HTML (.html, .htm) - Max 10MB
  • ✅ JSON (.json, .jsonl) - Max 10MB
Documents are uploaded to Google Cloud Storage and indexed by Vertex AI Search. Indexing typically takes 5-60 minutes depending on file count.
3

Upload

Click “Upload Document” and select your fileThe system will:
  1. Upload to Google Cloud Storage
  2. Create a Vertex AI Search DataStore
  3. Index the document (JSONL conversion)
  4. Generate a RAG tool ID (format: rag_{datastore_id})
4

Monitor Status

Watch the upload progress bar. Status will change:
  • Uploading → Indexing → Completed

​
Assigning RAG Tools to Agents

1

Edit Agent

Open the agent edit dialog and scroll to the RAG Tools section
2

Select DataStores

Check the boxes for the knowledge bases this agent should accessExample:
  • ✅ rag_product_documentation
  • ✅ rag_support_tickets_history
  • ☐ rag_marketing_materials
3

Save

The agent can now query these documents during conversations!

​
RAG Best Practices

  • Clean formatting: Remove unnecessary headers, footers
  • Consistent structure: Use headings, bullet points
  • Clear language: Avoid jargon unless necessary
  • Chunk size: Keep documents under 10MB for faster indexing
  • Separate by topic (product docs, support tickets, policies)
  • Separate by audience (internal vs customer-facing)
  • Agents can use multiple DataStores simultaneously
  • Include keywords in your documents
  • Use semantic search (Vertex AI Search handles this automatically)
  • Test queries with the agent to verify retrieval accuracy

​
🔌 MCP Integrations

MCP (Model Context Protocol) connects your agents to external platforms like GitHub, Firecrawl, and custom services.

​
Available Platforms

GitHub

Manage repos, issues, and PRs

Firecrawl

Web scraping and crawling
Supported platforms: GitHub, Firecrawl, and custom MCP servers. Easily extensible for any service with MCP implementation. See the MCP Integration Guide for details.

​
Setting Up MCP Credentials

Shared credentials for all team members
1

Navigate

Go to Team > MCP in the sidebar
2

Select Platform

Click “Connect” next to a platform (e.g., GitHub, Firecrawl)
3

Choose Auth Type

  • API Key: Enter the API key/token
  • OAuth: Click “Authorize” to connect via OAuth
Shared Access: All agents in your organization can now use this platform (if granted permission).

​
Enabling MCP for an Agent

1

Edit Agent

Open the agent edit dialog and scroll to the MCP Platforms section
2

Select Platforms

Check the boxes for platforms this agent should access:
  • ✅ GitHub
  • ✅ Firecrawl
  • ☐ Custom MCP (if configured)
3

Save

Click “Save” to enable MCP integration for this agent

​
🔑 API Keys

Generate API keys to integrate Junis agents into external applications.

​
Creating an API Key

1

Go to Dashboard

Navigate to Admin > Dashboard, then click the “API Keys” tab
2

Create Key

Click ”+ New API Key”
3

Configure

API Key Settings
Name: My Integration Key
Description: For mobile app integration
Permissions: Read/Write
Expiration: 90 days (optional)
4

Copy Key

⚠️ Important: The key is shown only once. Copy it immediately!Format: jns_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx (32 characters after prefix)

​
Using an API Key

Authentication Header: 
X-API-Key: jns_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Example Request (OpenAI-compatible): 
curl https://api.junis.ai/api/external/v1/chat/completions \
  -H "X-API-Key: jns_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "orchestrator",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
OpenAI SDK Compatible: Use the OpenAI Python SDK with base_url="https://api.junis.ai/api/external/v1" for seamless integration.

​
Managing API Keys

  • See all API keys for your organization
  • Check last used timestamp
  • View permissions and expiration

​
⚙️ Organization Settings

Admin/Owner Only: This section is only accessible to Admins and Super Admins.
Navigate to Admin > Dashboard > Settings tab to configure organization-wide options.

​
Team Settings

  • Name: Your organization’s display name
  • Slug: URL-friendly identifier (e.g., acme-corp)
  • Logo: Upload a logo image (PNG, JPG, max 2MB)
  • Description: Public description (shown on showcase page)
  • Enable Public Profile: Make your organization discoverable on junis.ai/showcase
  • Welcome Suggestions: Customize suggested prompts for visitors
  • Featured Agents: Select which agents to showcase publicly
  • Company Wallet: Ethereum wallet address for receiving agent usage payments
  • Default Agent Pricing: Set default price per use (USDC micro)
  • View current usage statistics
  • Set monthly limits (Pro/Enterprise plans)
  • Configure billing alerts

​
User Management

Invite Team Members:
1

Go to User Management

Navigate to Admin > Dashboard > Users tab
2

Click Invite

Click the ”+ Invite User” button
3

Enter Email

Type the email address of the person you want to invite
4

Assign Role

  • Member: Can use agents, view dashboard
  • Support: + Can view agent configurations
  • Admin: + Can create/edit agents, manage settings
5

Send Invitation

Click “Send Invite” - the user will receive an email invitation
Change User Roles:
  • Find the user in the list
  • Click the role dropdown
  • Select the new role
Remove Users:
  • Click the trash icon next to a user
  • Confirm removal

​
🔍 Advanced Features

​
Scheduled Tasks

Automate agent execution with schedules (cron jobs).
1

Navigate

Go to Admin > Schedules (if available in your plan)
2

New Schedule

Click ”+ New Schedule”
3

Configure

Schedule Configuration
Name: Daily Report Generator
Prompt: Generate today's summary report
Schedule: Daily at 9:00 AM
Timezone: America/New_York
Session: Create new session each time
4

Test

Click “Run Now” to test before saving

​
Workflow Builder (Coming Soon)

Visual drag-and-drop interface for building complex agent workflows. Features (planned):
  • Drag agents from a palette
  • Draw connections between agents
  • Set conditional logic (if/else branches)
  • Test workflows in real-time
  • Export/import workflow templates

​
💡 Tips for Effective Admin Management

  • Begin with a single LLM Agent to handle basic tasks
  • Test thoroughly before adding complexity
  • Add Sequential/Parallel agents only when you need multi-step workflows
  • Monitor performance before scaling up
Naming Convention:
  • orchestrator - Main router
  • customer_support_agent - Customer-facing
  • research_analyst - Data analysis
  • email_sender - Action agents
Groups:
  • Front-line: Customer support, sales
  • Back-office: Data processing, reporting
  • Specialized: Legal, technical documentation
Good:
You are a customer support agent for Acme Corp.

Your responsibilities:
1. Answer questions about products A, B, C
2. Help troubleshoot common issues (login, billing)
3. Escalate to human support for complex cases

Important: Be friendly, concise, and professional.
Always verify user identity before sharing account info.
Bad:
You are a helpful assistant.
  • Check Dashboard metrics weekly
  • Review Recent Activity to see how agents are being used
  • Update agent instructions based on user feedback
  • Test edge cases regularly
  • Rotate API keys every 90 days
  • Use user-level MCP credentials for personal platforms (e.g., personal GitHub)
  • Set payment thresholds for paid agents
  • Review user permissions quarterly
  • Enable 2FA for all admin users (if available)

​
🆘 Common Issues and Solutions

Possible Causes:
  • Agent is inactive (check the toggle)
  • Agent is not connected to Orchestrator (check relationships)
  • Model API key is missing (verify in SQLAdmin)
Solution:
  1. Check agent status (green toggle)
  2. Verify Orchestrator relationships
  3. Test with a simple message
  4. Check Dashboard > Recent Activity for errors
Possible Causes:
  • Document still indexing (wait 1-2 minutes)
  • RAG tool not assigned to agent
  • Query doesn’t match document content
Solution:
  1. Check DataStore status (should be “Completed”)
  2. Verify agent has the RAG tool enabled
  3. Test with a query that matches document keywords
  4. Check Vertex AI Search console for indexing errors
Possible Causes:
  • Credentials expired or invalid
  • Platform not enabled for agent
  • Rate limiting from external platform
Solution:
  1. Re-authenticate the MCP platform
  2. Check agent MCP platforms list
  3. Verify API key permissions on the external platform
  4. Check error logs in Dashboard > Recent Activity
Possible Causes:
  • Browser cache issue
  • Relationships not saved correctly
Solution:
  1. Hard refresh the page (Ctrl+Shift+R or Cmd+Shift+R)
  2. Switch to List view to verify relationships
  3. Re-save the relationships if needed

​
📖 Next Steps

MCP Integration Guide

Connect external platforms via Model Context Protocol

Custom Tools

Register Python functions as agent tools

API Reference

Integrate Junis into your applications

​
🎓 Admin Interface Certification

Congratulations! You now know how to:
✅ Navigate the Admin dashboard and interpret metrics ✅ Create, edit, and delete agents with confidence ✅ Manage agent relationships and build workflows ✅ Upload RAG documents and assign to agents ✅ Configure MCP integrations (org-level and user-level) ✅ Generate and manage API keys securely ✅ Configure organization settings and team permissions ✅ Troubleshoot common issues independently
Need More Help? Check the API Reference for programmatic management or contact us at contact@junis.ai.