Documentation v1.4

Plis AI Documentation

Complete guide to building AI agents, workflow automation, multi-agent teams, social media management, and enterprise features. 12+ LLM providers, 150+ workflow nodes, 10 social platforms.

Build AI Agents

Create custom AI agents with personality and capabilities.

Workflow Builder

150+ nodes for automations, CRM, cloud storage, and more.

Multi-Agent Teams

Build agent teams that collaborate on complex tasks.

Social Media Hub

Manage 10 platforms with unified inbox and AI responses.

Overview

Plis AI is a comprehensive, self-hosted AI assistant platform for building, deploying, and managing AI agents. Whether you need a simple chatbot, complex multi-agent workflows, or enterprise-grade automation, Plis AI provides the tools and infrastructure to make it happen.

Platform Highlights

Category Capabilities
LLM Providers 12+ providers: OpenAI, Anthropic, Google Gemini, DeepSeek, Grok, Azure, Mistral, Kimi, Qwen, Hugging Face, Ollama, LM Studio
Workflow Engine 150+ nodes including CRM (Salesforce, HubSpot, Zoho, Pipedrive), cloud storage (S3, GCS, Azure Blob), databases, and more
Agent Types Individual agents, Team agents (AgentTogether), Agent Mesh networks, Autonomous agents
Social Media 10 platforms: Twitter/X, LinkedIn, Facebook, Instagram, TikTok, Discord, Slack, Telegram, WhatsApp, Viber
Enterprise ComplianceGuard, Approval workflows, Audit logging, PII detection, Data retention policies
Analytics AgentPulse (performance metrics), Anomaly detection, Cost tracking, ROI analysis

Core Features

  • Multi-Provider LLM Support - Connect to 12+ AI providers including cloud and local models
  • Agent Studio - Visual interface for creating and configuring AI agents
  • MCP Integration - 80+ MCP server templates for extending capabilities
  • Workflow Builder - 150+ nodes for automation with visual drag-and-drop
  • Knowledge Base - RAG-powered document search and retrieval
  • Social Media Hub - Unified inbox, content studio, campaigns, and AI-powered responses

Advanced Agent Features

  • AgentTogether - Multi-agent teams with collaboration modes (Sequential, Parallel, Consensus)
  • Agent Mesh - Cross-platform agent communication with multiple protocols (A2A, MCP, OpenAI Functions)
  • AgentPulse - Real-time performance monitoring and anomaly detection
  • AgentRewind - Time travel debugging with execution traces and forking
  • AgentExchange - Marketplace for sharing and discovering agents
  • ContextGenius - Smart context compression with hierarchical summarization
  • SmartScheduler - Predictive task scheduling with pattern learning
  • VoiceFlow - Voice-triggered workflows and commands

Enterprise Features

  • Human-in-the-Loop - Approval nodes with configurable approval types and timeouts
  • Subflows - Nested workflows for reusability and composition
  • Browser Automation - Web scraping, clicking, typing, screenshots
  • Checkpoint & Retry - Save workflow state and resume from checkpoints
  • ComplianceGuard - PII detection, audit logging, content filtering
  • ConnectAnything - Create integrations from natural language descriptions

Quick Start

Get up and running with Plis AI in just a few steps:

1. Configure an AI Provider

Go to Settings > LLM Configuration and add at least one AI provider. You can use cloud providers like OpenAI or Anthropic, or run local models with Ollama.

2. Create Your First Agent

Navigate to Agents > Agent Studio and click "New Agent". Give it a name, select a model, and write a system prompt.

3. Start Chatting

Go to Chat, select your agent, and start a conversation!

AI Providers

Plis AI supports multiple AI providers, both cloud-based and local.

Provider Type Best For
OpenAI Cloud General purpose, GPT-4o, o1, o1-mini
Anthropic Cloud Coding, analysis, Claude Opus/Sonnet
Google Gemini Cloud Multimodal, large context, Gemini 2.0
Azure OpenAI Cloud Enterprise, compliance, gpt-4, o1, o3
DeepSeek Cloud Reasoning, deepseek-chat, deepseek-reasoner
Grok (xAI) Cloud Real-time data, grok-3, grok-2
Mistral Cloud Code, mistral-large, codestral
Kimi (Moonshot) Cloud Long context (128k), moonshot models
Alibaba Qwen Cloud Multilingual, qwen-max, qwen-plus
Hugging Face Cloud Open models, Llama-3.3, Mixtral
Ollama Local Privacy, offline, free, llama3
LM Studio Local GGUF models, desktop GUI
ⓘ
Getting API Keys
OpenAI: platform.openai.com/api-keys
Anthropic: console.anthropic.com/settings/keys
Google: aistudio.google.com/app/apikey
DeepSeek: platform.deepseek.com
Grok: console.x.ai
Mistral: console.mistral.ai

Local Providers (No API Key Required)

Ollama: Download from ollama.ai, then run ollama pull llama3 to download models.
LM Studio: Download from lmstudio.ai, download models in-app, and start the local server.

Agent Basics

An agent in Plis AI is a configured AI assistant with specific behavior, capabilities, and personality.

Agent Components

Component Description Required
Name Unique identifier for the agent Yes
Role Job title or expertise area No
System Prompt Behavior instructions Yes
Provider & Model AI service and model to use Yes
Tools Enabled capabilities (web search, etc.) No
MCP Servers External integrations No

Writing System Prompts

The system prompt is the most important part of your agent. It defines personality, expertise, and behavior.

Anatomy of a Good System Prompt

[Identity] + [Expertise] + [Guidelines] + [Format] + [Constraints]

Example System Prompt

System Prompt
You are PythonExpert, a senior Python developer with 10+ years of experience. Your expertise includes: 1. Core Python: Type hints, dataclasses, decorators, async/await 2. Web Frameworks: FastAPI, Flask, Django 3. Data Processing: Pandas, NumPy, data pipelines 4. Testing: pytest, TDD, mocking When writing code: - Always use type hints for function signatures - Follow PEP 8 style guidelines - Include docstrings for public functions - Handle errors explicitly, never use bare except Response format: 1. Brief explanation of the approach 2. Complete, runnable code with comments 3. Usage example 4. Potential improvements or alternatives Constraints: - Ask clarifying questions if requirements are unclear - Warn about security implications when relevant - Suggest tests for critical functionality
💡
Pro Tip
Be specific! "Be helpful" is too vague. Define exactly how the agent should respond, what format to use, and what boundaries to respect.

Tool Calling

Enable agents to perform actions beyond text generation.

Built-in Tools

Tool Description
Web Search Search the internet for current information
Code Interpreter Execute code for calculations and processing
HTTP Request Make API calls to external services
Read URL Extract and summarize webpage content
Message Agent Contact other agents in multi-agent workflows
⚠
Auto-Execute Caution
When auto-execute is enabled, tools run without user confirmation. Only enable this for trusted agents performing safe operations.

MCP Servers

Model Context Protocol (MCP) servers extend agent capabilities with external tools and services.

Popular Integrations

Server Purpose
GitHub Repos, issues, PRs, code search
Slack Send messages, read channels
PostgreSQL Query databases
Brave Search Web search with privacy
Filesystem Read/write local files

Configuration Example

JSON
{ "name": "github", "transport": "stdio", "command": "npx", "args": ["-y", "@modelcontextprotocol/server-github"], "env": { "GITHUB_TOKEN": "ghp_your_token_here" } }

Workflow Builder

Build automated processes with our visual drag-and-drop builder featuring 150+ nodes for AI, APIs, CRM, cloud storage, databases, and more.

How to Create a Workflow

  1. Navigate to Workflows

    Go to Workflows from the sidebar and click New Workflow.

  2. Add a Trigger

    Start with a trigger node: Manual, Schedule (cron), Webhook, Email, File, or RSS.

  3. Add Action Nodes

    Drag nodes from the palette: AI/LLM, HTTP Request, Email, Database, CRM, etc.

  4. Connect and Configure

    Draw connections between nodes and configure each node's settings.

  5. Test and Activate

    Use the Test button to run the workflow, then toggle to Active.

Workflow Triggers

Triggers determine when and how a workflow starts.

Trigger Description Use Case
Manual Run on-demand via button click Testing, one-time tasks
Schedule Run on cron schedule (daily, hourly, etc.) Reports, data sync, backups
Webhook Run when HTTP request is received External integrations, API callbacks
Email Run when email arrives (IMAP polling) Email processing, auto-replies
File Run when file changes in directory File processing, uploads
RSS Run when new RSS feed item appears Content monitoring, news alerts
GitHub Run on GitHub events (push, PR, issue) CI/CD, code review automation
GitLab Run on GitLab events (push, MR, pipeline, deployment) CI/CD, merge request automation, deployment workflows
Database Run on database changes (INSERT, UPDATE, DELETE) Data sync, notifications, audit trails
Form Run when form is submitted Lead capture, surveys, registrations
Message Run on chat messages (Slack, Discord, Teams) Chatbots, command handlers
API Poll Periodically check an API for changes Status monitoring, data refresh
Calendar Run on calendar events (Google, Outlook) Meeting prep, reminders, scheduling

GitHub Trigger Events

The GitHub trigger supports webhook events for repository automation:

Event Description
Push Triggered on git push to any branch
Pull Request Triggered on PR events (opened, closed, merged, synchronized, review requested)
Issues Triggered on issue events (opened, closed, edited, labeled, assigned)
Issue Comment Triggered on comments on issues and pull requests
Release Triggered when a release is published, created, edited, or deleted
Star Triggered when someone stars or unstars the repository
Fork Triggered when someone forks the repository
Watch Triggered when someone watches the repository
Workflow Run Triggered when a GitHub Actions workflow run completes
Deployment Triggered on deployment and deployment status events
Check Run Triggered when check run status changes
Branch/Tag Triggered when branches or tags are created or deleted

GitHub Actions (18+ operations)

  • Repository: Get repo info, list branches, list tags, get contents
  • Issues: Create, edit, close, reopen, list, add comment, add labels
  • Pull Requests: Create, edit, merge, close, list, add reviewers, add comment
  • Releases: Create, edit, delete, list, upload assets
  • Workflows: Trigger workflow, get run status, list runs, cancel run
  • Files: Get file, create/update file, delete file

GitLab Trigger Events

The GitLab trigger supports 12 event types for comprehensive CI/CD and DevOps automation:

Event Description
Push Triggered on git push to repository
Tag Triggered when tags are created or deleted
Merge Request Triggered on MR events (open, close, merge, update)
Issue Triggered on issue events (open, close, update, reopen)
Comment Triggered on comments (issues, MRs, commits, snippets)
Pipeline Triggered on pipeline status changes
Job Triggered on job status changes
Deployment Triggered on deployment events
Release Triggered on release events (create, update)
Wiki Page Triggered on wiki page events (create, update, delete)
Confidential Issue Triggered on confidential issue events
Confidential Comment Triggered on confidential note events

GitLab Actions (30+ operations)

  • Files: Create, edit, delete, get file, list files in repository
  • Issues: Create, edit, close, reopen, get issue, list issues, add comment
  • Merge Requests: Create, edit, merge, approve, get MR, list MRs, add comment
  • Pipelines: Trigger, cancel, retry pipeline, list pipelines, get/retry/cancel jobs
  • Deployments: List deployments, get deployment, list/create/stop environments
  • Releases: Create, update, delete, get release, list releases
💡
Webhook Security
Enable HMAC signature verification for webhooks to validate incoming requests. Configure the secret in the trigger settings.

150+ Workflow Nodes

Our workflow engine supports a wide variety of nodes comparable to n8n and Make.com.

AI/LLM Nodes

Node Providers
LLM OpenAI, Claude, Gemini, DeepSeek, Grok, Mistral, Ollama, LMStudio
Agent Use any configured Plis AI agent in workflows
Text-to-Speech OpenAI TTS, ElevenLabs, Azure Speech
Speech-to-Text OpenAI Whisper, Azure, Google Speech
Image Generation DALL-E, Stable Diffusion
Vision/OCR GPT-4 Vision, Claude Vision, Gemini Vision

CRM Integrations

CRM Operations
Salesforce Create, Read, Update, Delete, Query (SOQL), Search (SOSL), Describe
HubSpot Contacts, Companies, Deals, Tickets - full CRUD + search + associations
Zoho CRM Leads, Contacts, Accounts, Deals - full CRUD + search
Pipedrive Deals, Persons, Organizations, Activities - full CRUD + search

Cloud Storage

Service Operations
AWS S3 Upload, Download, List, Delete, Copy, Presigned URLs, Metadata
Google Cloud Storage Upload, Download, List, Delete, Signed URLs
Azure Blob Upload, Download, List, Delete, SAS URLs

Database Operations

Database Operations
PostgreSQL Query, Execute, Insert, Update, Delete, Upsert, Transactions, Stored Procedures
MySQL Full SQL operations with SSL/TLS support
SQLite Local database operations
MongoDB Find, Insert, Update, Delete, Aggregate
Redis Get, Set, Delete, Lists, Hashes, Pub/Sub

Other Node Categories

  • Communication: Email (SMTP), Slack, Discord, Telegram, WhatsApp, Viber
  • Data/API: HTTP Request (OAuth2, retries), GraphQL, JSON Transform, CSV Parser
  • Flow Control: Condition, Switch, Loop, Parallel, Wait, Merge, Split
  • Output: File Write, Google Sheets, Airtable, Notion, Response
  • Social Media: Twitter, Instagram, Facebook, LinkedIn, TikTok
  • DevOps: GitHub, GitLab, Docker, Kubernetes, SSH, Shell Command
  • Advanced: Approval, Human-in-Loop, Multi-Agent Orchestration, Memory

Team Agents (AgentTogether)

Enable multiple AI agents to work together on complex tasks through teams, collaboration sessions, task delegation, and consensus voting.

Team Configuration

Setting Description
Team Name Display name for the team
Team Type Sequential, Parallel, Hierarchical, Consensus, or Swarm
Share Memory Enable memory sharing between agents
Share Context Enable context sharing between agents
Context Retention How long to retain shared context (hours)
Allow Direct Communication Allow agents to communicate directly
Require Coordinator Approval Require approval for actions
Max Concurrent Tasks Maximum tasks running simultaneously

Agent Roles

  • Coordinator - Manages workflow and delegates tasks
  • Specialist - Expert in specific domain
  • Reviewer - Reviews and validates outputs
  • Executor - Executes defined tasks
  • Observer - Monitors without acting

Collaboration Modes

Mode Description
Sequential Agents work one after another in order
Parallel Agents work simultaneously on different tasks
Hierarchical Tree-based delegation from coordinator to specialists
Consensus Democratic decision-making with voting
Swarm Dynamic self-organization based on task needs

How to Create a Team

  1. Navigate to Team Agents

    Go to Agents > Team Agents from the sidebar.

  2. Create New Team

    Click Create Team and configure team settings (type, sharing, constraints).

  3. Add Team Members

    Add agents to the team and assign roles (Coordinator, Specialist, Reviewer, etc.).

  4. Start Collaboration Session

    Create a session with an objective and the team will work together to complete it.

Agent Mesh (Cross-Platform Protocol)

Connect Plis AI agents with external AI agents running on other platforms like OpenAI, Anthropic, Azure, or custom APIs.

Supported Platforms

  • PlisAI Internal - Connect your own Plis AI agents
  • OpenAI - Connect to GPT-based agents
  • Anthropic - Connect to Claude-based agents
  • Azure AI - Connect to Azure OpenAI agents
  • Custom - Connect to any custom agent API

Protocols

Protocol Description Use Case
A2A Agent-to-Agent native protocol PlisAI to PlisAI communication
MCP Model Context Protocol (JSON-RPC 2.0) Claude, tool-based agents
OpenAI Functions OpenAI function calling format GPT-based agents
Custom Your own protocol implementation Proprietary systems

Topology Types

Topology Description
Star All agents connect through a central coordinator
Mesh Every agent can communicate directly with every other agent
Hierarchical Tree structure with levels of authority
Ring Agents arranged in a circular communication pattern

Message Routing

  • Direct - Point-to-point messaging between agents
  • Coordinator - All messages route through coordinator
  • Broadcast - Messages sent to all connected agents

Trust Levels

Level Description
1 - Untrusted Minimal access, heavy monitoring
2 - Low Trust Limited actions, rate limited
3 - Medium Trust Standard access (default)
4 - High Trust Extended permissions
5 - Fully Trusted Full access, minimal restrictions

How to Set Up Agent Mesh

  1. Navigate to Agent Mesh

    Go to Agents > Agent Mesh from the sidebar.

  2. Register External Agents

    Click Register Agent and enter the agent's endpoint, protocol, and credentials.

  3. Create Mesh Network

    Click Create Network and select topology type and routing mode.

  4. Establish Connections

    Add agents to the network and configure trust levels for each connection.

AgentPulse (Performance Analytics)

Real-time agent performance monitoring with cost tracking, anomaly detection, and ROI analysis.

Metrics Tracked

Metric Description
Success Rate Percentage of successful task completions
Latency Response time metrics (p50, p95, p99)
Token Usage Input/output token consumption
Cost Estimated spending per agent
Error Rate Failed request percentage
Health Score Overall agent health (0-100)

Anomaly Detection

AgentPulse automatically detects anomalies in agent behavior:

  • Success Rate Drop - Success rate below baseline
  • Latency Spike - Response time above threshold
  • Cost Spike - Unexpected cost increase
  • Volume Spike - Unusual task volume
  • Error Pattern - Recurring error patterns
  • Model Drift - Model behavior changes over time

Performance Scorecard

View an agent's health scorecard in Agents > Agent Details > Pulse tab. The scorecard shows current metrics compared to baseline, with color-coded indicators.

💡
Set Up Baselines
Click "Calculate Baseline" after at least 7 days of agent activity to enable accurate anomaly detection.

AgentRewind (Time Travel Debugging)

Interactive debugging system that captures every decision point during agent execution, allowing you to replay, inspect, and fork agent sessions.

Features

  • Execution Traces - Captures every decision point, model call, and tool execution
  • State Snapshots - Full context snapshots at each step
  • Forking - Fork from any trace point with modified context
  • What-If Analysis - Test alternative scenarios from any decision point
  • Run Comparison - Compare two runs side by side
  • Token Tracking - Token and latency tracking per step

Trace Types

Type Description
decision Agent decision points with alternatives considered
model_call LLM API calls with prompts and responses
tool_call Tool executions with inputs and outputs
error Errors and exceptions that occurred

How to Use AgentRewind

  1. Enable Tracing

    Enable execution tracing in the agent's settings.

  2. Run Agent Tasks

    Use the agent normally. All executions are automatically traced.

  3. Open AgentRewind

    Go to Agents > AgentRewind and select an execution run.

  4. Browse Timeline

    View the timeline of traces. Click any point to see the full state snapshot.

  5. Fork and Explore

    Click Fork from here to create a new run from that point with modified context.

AgentExchange (Marketplace)

Marketplace for sharing, discovering, and installing AI agents and workflows.

Features

  • Publish - Share agents and workflows
  • Discover - Browse by category, search, or trending
  • Install - One-click installation
  • Reviews - Community ratings and reviews
  • Analytics - Installation and usage tracking
  • Publisher Profiles - Follow your favorite publishers

Categories

  • Productivity
  • Data Analysis
  • Content Creation
  • Customer Service
  • Development
  • Marketing
  • Finance
  • HR & Recruitment
  • Research

How to Install an Agent

  1. Go to Agents > AgentExchange
  2. Browse or search for agents
  3. Click on an agent to view details and reviews
  4. Click Install to add to your account

How to Publish

  1. Create and test your agent or workflow
  2. Click Publish from the agent/workflow details
  3. Fill in listing details (name, description, category)
  4. Add screenshots and documentation
  5. Submit for review

Enterprise Workflow Features

Advanced workflow capabilities for enterprise-grade automation.

Human-in-the-Loop Approval Node

Pause workflow execution to require human approval before continuing.

Setting Description
Approval Type Any (first approver), All (unanimous), Majority (>50%), Threshold (specific count)
Timeout Hours before auto-action (reject, approve, escalate, fail)
Notifications Email, Slack, in-app, webhook notifications to approvers
Comments Optional comments with required comment on rejection

Subflow/Nested Workflow Node

Execute another workflow as a nested subflow for workflow reusability.

  • Execution Modes - Sync, async, or parallel
  • Input/Output Mapping - Map data between parent and child workflows
  • Context Inheritance - Pass context to subflows
  • Error Isolation - Prevent subflow failures from affecting parent
  • Recursion Protection - Configurable max depth

Browser Automation Node

Automate web browser actions including navigation, clicking, typing, scraping, and screenshots.

Action Description
Navigate Go to URL with configurable wait conditions
Click Click elements by CSS selector, XPath, or text
Type Enter text into form fields
Scrape Extract data using JSON-based scraping rules
Screenshot Capture full page, viewport, or specific elements
Execute Script Run custom JavaScript in page context

Features: Headless mode, stealth mode, custom viewport, cookie injection, proxy support.

Checkpoint & Retry Node

Save workflow state and enable retry/resume capabilities.

  • Checkpoint - Save state at any point with configurable TTL
  • Resume - Resume from any saved checkpoint
  • Retry Block - Automatic retry with backoff strategies (fixed, linear, exponential)
  • Fallback Values - Define fallback on final failure

Real-time Execution Visualization

Watch workflows execute in real-time with live node status updates via Server-Sent Events (SSE). See node progress, data flow between nodes, and approval requests as they happen.

ComplianceGuard (Governance)

Automated compliance monitoring, PII detection, audit logging, and governance policy enforcement.

Features

Feature Description
Rule Engine Define and evaluate compliance rules
PII Detection Automatic detection of sensitive data patterns
Audit Logging Comprehensive action logging for compliance
Data Retention Policy-based data lifecycle management
Content Filtering Block or flag prohibited content
Compliance Reports Generate regulatory compliance reports

Supported PII Patterns

  • Email addresses
  • Phone numbers
  • Social Security Numbers
  • Credit card numbers
  • IP addresses
  • Custom patterns (regex)

Rule Types

  • Content Filter - Block specific content patterns
  • PII Detection - Mask personal identifiable information
  • Data Retention - Auto-delete after specified period
  • Rate Limiting - Prevent abuse

How to Set Up

  1. Go to Settings > ComplianceGuard
  2. Create compliance rules based on your requirements
  3. Configure PII detection patterns to scan
  4. Set up data retention policies
  5. Enable audit logging for required actions
  6. Generate compliance reports as needed

ConnectAnything (Natural Language Integrations)

AI-powered system that creates integrations from natural language descriptions. Simply describe what you want to connect and Plis AI builds the integration.

Features

  • Natural Language Input - Describe integrations in plain English
  • Auto-Detection - Automatically detect known services
  • Endpoint Discovery - Discover and configure endpoints
  • Multi-Auth - OAuth2, API key, Bearer, Basic auth support
  • Test & Validate - Test integrations before activation

Pre-configured Services

Service Category
Shopify Ecommerce
Stripe Payments
Notion Productivity
Slack Communication
HubSpot CRM
Airtable Database
GitHub Development
Google Sheets Productivity

Example Usage

"Connect to my Shopify store to manage products and orders"

Plis AI will:

  1. Detect Shopify service
  2. Configure base URL for your store
  3. Set up OAuth2 authentication
  4. Discover available endpoints (products, orders, customers)
  5. Guide you through authorization

How to Use

  1. Go to Integrations > ConnectAnything
  2. Describe the integration you want in natural language
  3. Review the generated configuration
  4. Complete authentication if required
  5. Test the integration
  6. Activate when ready

SmartScheduler (Predictive Scheduling)

AI-powered task scheduling that learns your patterns and proactively suggests optimal execution times.

Features

  • Pattern Learning - Learn time-based usage patterns
  • Task Prediction - Predict upcoming tasks with confidence scoring
  • Proactive Suggestions - AI-generated task suggestions
  • Smart Queue - Intelligent task prioritization
  • Availability Tracking - User availability windows
  • Optimal Timing - Calculate optimal execution times

Pattern Types

Type Description Example
Time-based Regular time patterns Daily standup at 9am
Event-based Triggered by events PR merged triggers deploy
Sequence-based Task sequences Build → Test → Deploy

How to Use

  1. Go to Settings > SmartScheduler
  2. Allow pattern learning (7-30 days of data recommended)
  3. Set your availability windows
  4. View learned patterns and predictions
  5. Accept or dismiss suggestions as they appear
💡
Better Predictions
The more you use Plis AI, the better SmartScheduler becomes at predicting your needs. Allow at least 2-4 weeks for optimal pattern learning.

Knowledge Base

Upload documents and let agents search them using RAG (Retrieval-Augmented Generation).

Supported Formats

  • PDF documents
  • Word documents (.docx)
  • Text files (.txt, .md)
  • Web pages (via URL)

How to Use Knowledge Base

  1. Create a Collection

    Go to Knowledge Base and click New Collection. Give it a name and description.

  2. Upload Documents

    Click Add Documents and upload files or paste URLs. Documents are automatically chunked.

  3. Attach to Agent

    In Agent Studio, enable Knowledge Base and select your collection(s).

  4. Query via Chat

    Ask questions in chat - the agent will search and cite relevant documents.

Agent Templates

Start quickly with pre-built agent templates for common use cases.

Available Templates

  • Python Developer Pro - Expert Python development with best practices
  • DevOps Engineer - Infrastructure, CI/CD, and cloud deployment
  • Security Analyst - Security assessment and vulnerability analysis
  • Technical Writer - Documentation and technical content
  • Code Reviewer - Comprehensive code review and quality assessment
  • Customer Support - Friendly, helpful customer assistance
  • Research Assistant - Information gathering and synthesis

Access templates in Agents > Agent Studio when creating a new agent.

Social Media Hub - Account Setup

Connect your social media and messaging accounts to manage them all from one place.

Supported Platforms

Platform Auth Type Features
WhatsApp Business API Token Messages, templates, media, location, interactive
Viber Bot Token Messages, rich media, keyboards, carousel
Telegram Bot Token Messages, photos, documents
Discord Bot Token Messages, embeds, reactions
Slack Bot Token / OAuth Messages, channels, files
Twitter/X OAuth 2.0 Tweets, replies, media
Instagram OAuth (Meta) Posts, stories, comments
Facebook OAuth (Meta) Page posts, comments, messages
LinkedIn OAuth 2.0 Posts, shares
TikTok OAuth 2.0 Video posts

How to Connect an Account

  1. Go to Social Media Hub

    Navigate to Social Media Hub > Accounts tab.

  2. Choose Platform

    For OAuth platforms, click the platform card to start authentication. For bot platforms, select from the dropdown.

  3. Enter Credentials

    For messaging platforms: Enter your API token and platform-specific fields (Phone ID for WhatsApp, etc.).
    For OAuth platforms: Enter Client ID and Secret from the developer portal.

  4. Verify and Connect

    Click Connect Platform. The token will be verified and your account connected.

ⓘ
Getting API Credentials
Each platform setup section includes step-by-step instructions for obtaining your API tokens from the developer portals.

Unified Inbox

View and respond to conversations from all connected messaging platforms in one place.

Features

  • Unified View - All WhatsApp, Viber, Telegram conversations in one inbox
  • Status Filters - Filter by Active, Pending, Resolved, Escalated
  • Platform Filters - Show conversations from specific platforms
  • Real-time Chat - Send and receive messages instantly
  • Unread Badges - See unread count at a glance
  • Contact Info - View contact name, phone, avatar, and metadata

How to Use the Inbox

  1. Go to Social Media Hub > Inbox tab
  2. Select a conversation from the left sidebar
  3. View message history in the chat area
  4. Type your response and click Send (or press Enter)
  5. Use the toolbar to: Assign agent, Resolve conversation, View contact details
💡
Assign AI Agents
Click the gear icon in the chat header to assign an AI agent for automated responses. Configure handoff keywords to transfer to humans when needed.

Content Studio

Create, schedule, and publish content across multiple platforms with AI-powered writing tools.

Features

  • Multi-Platform Posting - Publish to multiple accounts at once
  • Media Upload - Attach images and videos to posts
  • Scheduling - Schedule posts for later or publish immediately
  • AI Caption Generator - Generate captions with different tones (professional, casual, funny, etc.)
  • Hashtag Suggestions - AI-powered hashtag recommendations
  • Content Rewriter - Make content shorter, longer, formal, or more engaging
  • Translation - Translate content to 9+ languages
  • Drafts - Save work in progress and return later

How to Create a Post

  1. Go to Content Studio

    Navigate to Social Media Hub > Content Studio tab.

  2. Select Platforms

    Check the platforms you want to post to from your connected accounts.

  3. Write Content

    Type your post content. Use AI tools on the right to generate captions, suggest hashtags, or rewrite.

  4. Add Media (Optional)

    Click the media upload area to add images or videos.

  5. Publish or Schedule

    Click Publish to post now, or choose Schedule for Later and pick a date/time.

AI Tools

Tool Description
Generate Caption Describe your post and select a tone. AI creates an engaging caption.
Hashtag Suggestions Enter a topic to get relevant hashtag recommendations. Click to add.
Rewrite Content Transform your content: Make Shorter, Make Longer, More Formal, More Engaging.
Translate Translate content to Spanish, French, German, Japanese, Chinese, and more.

Marketing Campaigns

Create and manage marketing campaigns with bulk messaging and scheduling capabilities.

Features

  • Campaign Types - Broadcast (one-time), Drip (multi-step), Promotional, Newsletter
  • Audience Targeting - All contacts, active subscribers, new contacts, re-engagement
  • Quick Broadcast - Send one-time messages to selected audiences
  • Personalization - Use and other variables in messages
  • Tracking - Track opens, clicks, and engagement
  • Quiet Hours - Respect 9PM-9AM quiet hours option
  • Campaign Analytics - View messages sent, reach, and engagement rate

How to Create a Campaign

  1. Go to Campaigns

    Navigate to Social Media Hub > Campaigns tab and click New Campaign.

  2. Configure Campaign

    Enter name, select type (Broadcast, Drip, etc.), choose target platforms and audience.

  3. Write Message

    Write your campaign message. Use for personalization.

  4. Schedule

    Set start date and optional end date. Enable tracking options as needed.

  5. Launch

    Click Launch Campaign or Save as Draft to finish later.

Quick Broadcast

For simple one-time messages, use the Quick Broadcast section at the bottom of the Campaigns tab. Select platforms, choose audience, write message, and send immediately or schedule.

Agent Routing

Assign AI agents to your messaging accounts for automated customer support and responses.

Features

  • Auto-Respond - Agent automatically replies to incoming messages
  • Approval Mode - Review agent responses before sending
  • Handoff Keywords - Trigger transfer to human on specific words (e.g., "human", "agent", "help")
  • Priority Routing - Higher priority agents handle messages first
  • Performance Stats - Track conversations handled, messages sent, handoffs

How to Assign an Agent

  1. Go to Agent Routing

    Navigate to Social Media Hub > Agent Routing tab.

  2. Click Assign Agent

    Click the Assign Agent button to open the assignment dialog.

  3. Select Account and Agent

    Choose the social account (e.g., WhatsApp Business) and the AI agent to assign.

  4. Configure Settings

    Enable/disable auto-respond, set approval requirements, configure handoff keywords.

  5. Save Assignment

    Click Assign Agent to activate. Toggle the assignment on/off anytime.

⚠
Test Before Production
Always test agent responses with the approval mode enabled first. Review responses for quality before enabling full auto-respond.

Automation Rules

Create rules to automatically respond, tag, or escalate conversations based on triggers.

Trigger Types

Trigger Description Example
Keyword Message contains specific words "pricing", "refund", "hours"
New Conversation A new conversation starts Send welcome message
No Response No reply after X minutes Follow-up message
Business Hours Outside business hours Out-of-office reply
Sentiment Negative sentiment detected Escalate to manager

Action Types

  • Auto Reply - Send a predefined response
  • Assign Agent - Route to a specific AI agent
  • Tag Conversation - Add tags for organization
  • Send Template - Send a WhatsApp Business template
  • Escalate - Transfer to human agent with priority

Creating a Rule

  1. Go to Social Media Hub > Automation tab
  2. Click Create Rule
  3. Enter rule name and description
  4. Select trigger type and configure conditions
  5. Select action type and configure response
  6. Set priority (higher priority rules run first)
  7. Enable "Stop on match" if no other rules should run after this one
  8. Click Create Rule to save

Voice Flow

Control Plis AI with voice commands. Create triggers, shortcuts, and automate tasks hands-free.

Features

  • Voice Triggers - Define phrases that activate specific actions
  • Shortcuts - Quick voice commands for common tasks
  • Command History - View and analyze past voice commands
  • Multi-language - Support for multiple languages
  • Confidence Scoring - See how accurately commands were recognized

How to Create a Voice Trigger

  1. Go to Voice Flow

    Navigate to Voice Flow from the sidebar.

  2. Create New Trigger

    Click New Trigger and enter a name.

  3. Define Trigger Phrases

    Add phrases that activate this trigger (e.g., "start meeting", "begin meeting", "let's start").

  4. Configure Action

    Select what happens when triggered: Run workflow, send message, execute command, etc.

  5. Test and Enable

    Test the trigger and toggle it to active.

Voice Shortcuts

Shortcuts are quick voice commands for common actions. Go to Voice Flow > Shortcuts to manage them. Examples: "Read messages", "Check calendar", "Send report".

Context Genius

Intelligent context window management with LLM-powered summarization, conversation branching, and hierarchical compression.

Features

  • LLM-Powered Summarization - Intelligent message summarization using AI
  • Key Facts Extraction - Extract facts, decisions, and user preferences
  • Hierarchical Compression - Recent messages stay detailed, older ones are compressed
  • Conversation Branching - Create branches for "what-if" exploration
  • Content Pinning - Pin important content to preserve it
  • Cross-Branch Retrieval - Access context across conversation branches

Summarization Strategy

Recency Treatment
< 1 hour Full messages (no compression)
1-24 hours Light summarization (70% retention)
1-7 days Medium summarization (40% retention)
> 7 days Heavy summarization (20% retention)

How to Use

  1. Go to Context > Context Genius
  2. Select a conversation to manage
  3. View and manage summaries for different time periods
  4. Create branches to explore alternative conversation paths
  5. Pin important content you want to always preserve
💡
Enable per Agent
Context Genius can be enabled/disabled per agent in Agent Studio settings. Enable it for agents that need long conversation memory.

Analytics Dashboard

Monitor AI usage, costs, and performance across all your agents and workflows.

Metrics Tracked

Metric Description
Total Requests Number of API calls made to AI providers
Total Tokens Input + output tokens consumed
Estimated Cost Calculated cost based on provider pricing
Avg Latency Average response time in milliseconds
Tokens/Second Generation speed metric

Dashboard Views

  • Overview - Summary stats for selected time period
  • Usage Over Time - Chart showing usage trends
  • By Provider - Breakdown by AI provider (OpenAI, Claude, etc.)
  • By Agent - Usage per agent
  • By Model - Comparison across different models

Exporting Data

Click the Export CSV button to download usage data for external analysis or reporting.

Widget & Embed

Embed AI chat widgets on your website to let visitors interact with your agents.

Features

  • Customizable Design - Match your brand colors and style
  • Agent Selection - Choose which agent powers the widget
  • Domain Restrictions - Limit which domains can use the widget
  • Rate Limiting - Control usage to manage costs
  • Anonymous or Authenticated - Support both modes

How to Create a Widget

  1. Go to Widget Manager

    Navigate to Widget from the sidebar.

  2. Create API Key

    Click New API Key, give it a name, and select the agent to use.

  3. Configure Settings

    Set allowed domains, rate limits, and customize appearance.

  4. Copy Embed Code

    Copy the generated JavaScript snippet.

  5. Add to Website

    Paste the snippet before the closing </body> tag on your website.

Embed Code Example

HTML
<script src="https://your-plis-instance.com/widget.js" data-api-key="your-api-key-here" data-position="bottom-right" data-primary-color="#6366f1"> </script>
⚠
Keep API Keys Secret
While widget API keys are visible in page source, use domain restrictions and rate limits to prevent abuse. Never use these keys for server-side API calls.

Contact Support