> ## Documentation Index
> Fetch the complete documentation index at: https://docs.praxis-ai.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Welcome to Praxis AI
> The AI middleware that turns any AI model into a personalized Digital Twin — with your voice, your knowledge, and your tools. Deploy everywhere: LMS, web, mobile, voice, or API.
***
## Human-First in an AI-First World
Praxis AI amplifies human expertise through the creation of custom-built **Human-First Digital Twins**. These are not generic chatbots; they are sophisticated extensions of your personal and institutional knowledge. By making your expertise instantly available 24/7, we ensure that while the technology is advanced, the approach remains entirely human-centric — putting your people and their insights at the heart of every interaction.
Built as **AI middleware**, Praxis sits between your organization and the AI models you choose. You bring the knowledge, personality, and branding. We handle the infrastructure — from voice cloning to document intelligence, from Google Workspace automation to LMS integration.
**Already using ChatGPT, Claude, or Gemini?** Praxis doesn't replace them — it gives you **100+ models from every major provider**, all in one place, all trained on *your* content. Switch models anytime. No vendor lock-in. Ever.
***
## Who Would Benefit from Digital Twins?
Praxis AI Digital Twins are for anyone whose expertise is worth sharing — at scale.
Whether you're an educator, healthcare provider, legal professional, corporate trainer, or industry thought leader, a Digital Twin puts your knowledge, voice, and insights to work 24/7 — even when you can't be in the room. Designed for professionals who value both impact and integrity, our platform lets you amplify what makes you uniquely brilliant without sacrificing authenticity or control.
If you've ever wished you could be in two places at once, your Digital Twin is the next best thing. **High tech. Human touch. Always you.**
***
## Where To Start
Head to the Digital Twin login page at [pria.praxislxp.com/login](https://pria.praxislxp.com/login) and follow the prompts to create your Digital Twin.
Once you're in, customize your Digital Twin by uploading files — click the **+** sign next to the dialogue bar to open the Action Menu.
For a step-by-step walkthrough, see [Getting Started with Your Digital Twin](/mdx/user-guide/digital-twin-getting-started).
***
## Why Digital Twins?
The question isn't whether AI can help your organization. It's whether your AI actually *knows* your organization.
Upload documents, manuals, course materials, or policies to the **IP Vault**. Your Digital Twin answers from *your* knowledge — not generic internet content. Retrieval-Augmented Generation (RAG) ensures every response is grounded in your source material.
Give your Digital Twin a **cloned voice** powered by ElevenLabs. Hold real-time voice conversations in many languages with a Twin that doesn't just think like your organization — it *sounds* like it too.
A Digital Twin is available 24/7, in every timezone, in any language. Whether a student needs tutoring at midnight or an employee needs onboarding guidance on a weekend — your Twin is ready.
***
## Choose Any AI. Keep All of Them.
Praxis is the only platform that gives you **unified access to 100+ AI models** across every major provider — through a single interface, a single API, and a single knowledge base.
Claude (Opus, Sonnet, Haiku), Amazon Nova, Meta Llama 4, Mistral, Cohere Command — all accessed through your own AWS account with full data sovereignty.
GPT-5.2, GPT-5.1, GPT-5-mini, o4-mini, GPT-Image-1.5, Whisper for transcription, and real-time voice via GPT-Realtime.
Claude Opus 4.6, Sonnet, Haiku — direct API access with extended thinking, prompt caching, and 1M token context windows.
Gemini 3 Pro, 3 Flash, 2.5 Pro — with native thinking configuration, context caching, and multimodal capabilities.
Mistral Large, Pixtral Large, Magistral (reasoning), Codestral (code), Devstral, Voxtral (voice), and specialized embedding models — all via native SDK integration.
Grok-4, Grok-4 Fast Reasoning, and Grok Code Fast — with native integration for streaming, tool calling, vision, and real-time voice.
**No vendor lock-in.** Switch models per conversation, per assistant, or per institution. Your knowledge base, tools, and configurations stay the same regardless of which AI powers the response.
***
## Your Twin Works Where You Work
Digital Twins don't live in a silo. They connect to the tools your team already uses — and act on them.
**Gmail, Drive, Calendar, Sheets, Docs, Slides, Classroom, Meet, Maps** — your Twin reads emails, schedules meetings, searches files, grades assignments, and manages documents. All through natural conversation.
Native **LTI 1.3** integration with **Canvas, Moodle, Blackboard, Brightspace, and Schoology**. Students and faculty access Digital Twins seamlessly from within their LMS — single sign-on, course context, and role-based access included.
Embed Twins in any **web application** with the JavaScript SDK. Access them on the go with the **Pria phone app**. Or build entirely custom experiences with the REST API.
Expose your Twin as an **MCP server** for Claude Desktop, Cursor, or any MCP-compatible client. Your Twin's knowledge becomes accessible to every tool in your AI workflow.
Drop-in **OpenAI-compatible** endpoint. Any app that works with the OpenAI SDK works with your Digital Twin — instantly.
Connect Digital Twins together. A sales Twin can query the product knowledge Twin. A student Twin can tap into the research Twin. No data duplication — live, real-time queries.
***
## Talk to Your Twin
Four world-class voice providers — OpenAI GPT-Realtime, ElevenLabs, Gemini Live, and xAI Grok — plus talking-avatar options. Real-time conversations in many languages. Tool execution mid-conversation. And yes — your Twin can sound exactly like you.
Ultra-low-latency voice with 10+ built-in voices, full tool calling during conversations, and MCP server access. Your Twin can search documents, check calendars, and send emails — all while you talk.
Clone any voice for a branded experience. Your Digital Twin speaks with the voice of your CEO, your professor, your brand ambassador. Dynamic context injection ensures every response is personalized.
Google's native multimodal audio with 30 built-in voices, real-time tool calling, and noise reduction.
***
## What Can Your Digital Twin Do?
Upload documents, manuals, research papers, or course materials to the **IP Vault**. Your Twin uses Retrieval-Augmented Generation to answer from *your* content — with citations and source links.
Every file has **three access tiers** — Public, Private (institution members only), or Confidential. Confidential locks the document itself: other members can't open, download, or see its details (they see a locked marker instead), though its knowledge can still inform shared answers. HMAC-signed URLs provide time-limited access that expires in seconds, not hours.
```text wrap theme={null}
- "What does our company policy say about remote work?"
- "Summarize chapter 5 of the textbook I uploaded"
```
Send emails, schedule meetings, search Drive files, read spreadsheets, access Classroom data — all through conversation. Your Twin handles the complexity of Google APIs so you don't have to.
```text wrap theme={null}
- "Send an email to the marketing team about the launch date"
- "What's on my calendar this Thursday?"
- "Find the Q4 report in my Google Drive"
```
Web search, academic paper search (Google Scholar, PubMed, NIH), patent databases, and Wolfram Alpha computations — all available as built-in tools your Twin can use mid-conversation.
```text wrap theme={null}
- "Find recent research on CRISPR gene therapy published this year"
- "Calculate the compound interest on a $50,000 investment over 10 years"
```
Generate images with DALL-E 3, Amazon Nova Canvas, or Gemini native image generation. Analyze uploaded images, charts, or screenshots with vision models. Transcribe audio and video files — with speaker identification.
```text wrap theme={null}
- "Generate a hero image for our sustainability landing page"
- "Transcribe this meeting recording and identify who said what"
```
**Personal Memory** learns your preferences, projects, and working style across sessions and devices. **Shared Memory** creates a team-wide knowledge base that every user of a Twin can contribute to and access.
Memory is scoped by design — personal memories stay personal, shared memories stay within your institution. Users in one Twin never see memories from another, and institution boundaries are enforced at the database level.
```text wrap theme={null}
- "Remember that I prefer bullet-point summaries"
- "Save this project brief to shared memory so the whole team can access it"
```
For **Canvas** users: grade assignments, monitor class health, identify at-risk students, analyze participation — all through pre-built assistants or natural conversation with your Twin.
```text wrap theme={null}
- "Grade the latest batch of essay submissions using the rubric"
- "Which students haven't submitted their assignments this week?"
```
Your Digital Twin defines *who* your AI is — its personality, knowledge, and voice. **Assistants** define *what it can do*. Each assistant is a specialized skill that orchestrates Pria's **39 built-in tools** — web search, Google Scholar, Wolfram Alpha, DALL-E, Canvas LMS, Google Workspace, memory management, code execution, and more — into repeatable, agentic workflows.
Create assistants for any task: a **Research Assistant** that searches scholarly databases, synthesizes findings, and formats citations. A **Report Generator** that pulls data from your Vault and builds structured summaries. A **Socratic Tutor** that guides students through problems step by step. Each one runs autonomously, chaining tools together to deliver results — not just answers.
```text wrap theme={null}
- "Use the Research Assistant to find peer-reviewed sources on climate adaptation"
- "Run the Course Syllabus Generator for my new biology class"
- "Have the Data Analyst pull last quarter's metrics and build a summary"
```
Security is not a bolt-on — it's the foundation. Every message is encrypted with **AES-256 at rest** and **TLS 1.3 in transit**. An AI-powered **content moderation** engine screens user messages before they reach the model, blocking policy violations and instantly alerting administrators. Voice sessions use **ephemeral tokens** that expire after each session — no long-lived API keys ever reach the browser.
Praxis is **FERPA**, **COPPA**, and **SOC 2** compliant. Student data is used exclusively for educational service delivery — never for advertising, behavioral profiling, or training general-purpose AI models. Data remains the exclusive property of the institution, with written destruction certification upon contract termination.
See the [Privacy & Educational Consent](/mdx/introduction/privacy-edu-consent) policy for the full data governance framework, including 30-day deletion guarantees and NIST-approved data destruction.
Administrators manage permissions through **31 granular entitlements** across 9 modules — institutions, users, sessions, histories, feedbacks, assistants, AI models, payments, and SSO. Every admin action goes through a **3-layer authorization check**: JWT authentication, account-type gate, and institution-scoped resource access.
Control what each Twin can do: enable or disable individual tools, restrict which AI models are available, limit file uploads to specific users, toggle assistants on or off, and enforce membership status (Active, Pending, Suspended) that takes effect immediately — no session refresh required.
IP Vault files use a **three-tier access model** — Public, Private (institution members), or Confidential (owner only). Guest UI mode strips the interface to a focused conversation experience for public-facing deployments. Admins can even disable clipboard copying for sensitive environments.
See the [Entitlements guide](/mdx/admin-guide/entitlements) for the full permission matrix and preset configurations.
Bring your own API keys from **any provider** — OpenAI, Anthropic, Google, Mistral, xAI, Cohere, or AWS Bedrock. A **3-tier key resolution** system checks the custom model first, then the institution key, then the environment fallback — so you control exactly where your data flows.
With **Amazon Bedrock**, your data never leaves your own AWS account. IAM policies enforce least-privilege access, and your encryption keys stay under your control. Switch providers at any time — your knowledge base, assistants, tools, and conversation history stay the same. **Zero vendor lock-in.**
Credentials are encrypted at rest, excluded from API responses by default, and never logged. OAuth tokens refresh automatically, and HMAC-signed URLs use time-safe comparison to prevent timing attacks.
See [Bring Your Own Technology](/mdx/introduction/byot) for the full BYOT architecture and provider comparison.
Every conversation is logged as a reviewable transcript — admins with the right entitlements can search, filter, and analyze interactions across their institution. Session tracking records user activity, login timestamps, and engagement patterns. Conversations are soft-deleted (never hard-removed), preserving a complete audit trail for compliance reviews.
The **credit system** gives administrators real-time visibility into AI consumption. Choose between institution-pooled credits (shared budget) or personal allocations (per-user limits). Set low-credit alerts, auto-transfer thresholds, and per-user credit awards. Cached tokens are charged at 50% to optimize costs automatically.
See [Administration & Analytics](/mdx/admin-guide/introduction) for the full suite of monitoring and governance tools.
***
## Platform at a Glance
Create, manage, and switch between Digital Twins from a single dashboard. Each Twin has its own personality, knowledge base, tools, and branding. Invite users, manage access, and track usage.
Real-time speech-to-speech in **many languages** with four voice providers: **OpenAI GPT-Realtime** for low-latency conversations with full tool and MCP support, **Gemini Live** for native multimodal audio with real-time tool calling, **xAI Grok** for Grok-powered voice with built-in voices, or **ElevenLabs** for custom voice clones and branded experiences. See [Convo Mode](/mdx/user-guide/interface/convo-speech) for details.
Embed Digital Twins directly in **Canvas, Moodle, Blackboard, Brightspace, or Schoology** via LTI 1.3. Students and faculty access the Twin without leaving their LMS. Course context, roles, and single sign-on are handled automatically.
Embed a Digital Twin as a **voice-enabled support agent** directly in your web application. Using the [Web SDK](/mdx/integrations/web/introduction), your Twin runs in [Convo Mode](/mdx/user-guide/interface/convo-speech) — visitors get real-time voice conversations, tool execution, and RAG-powered answers without leaving your site.
Full REST API with **OpenAPI/Swagger documentation**, a JavaScript SDK for web embedding, an OpenAI-compatible Chat Completions endpoint, and MCP server support. Build anything.
***
## Built for Every Team
Deploy AI tutors that speak your students' language — literally. LTI integration, Canvas tools, assignment grading, class health monitoring, and FERPA-compliant data handling. Students get personalized coaching 24/7 without adding to faculty workload.
Scale institutional knowledge without scaling headcount. White-label branding, multi-tenant architecture, role-based access, and full audit trails. Onboard employees, automate support, and centralize expertise in Digital Twins that never forget.
REST APIs, JavaScript SDK, OpenAI-compatible Chat Completions, and MCP protocol support. Embed Digital Twins in any application, connect them to any AI workflow, or build entirely new experiences on top of the platform.
Voice-enabled Digital Twins with cloned voices for branded customer interactions. Real-time tool execution during conversations. Your Twin doesn't just answer questions — it takes action on behalf of your customers.
***
## Enterprise-Grade From Day One
Speak, read, and interact in any language — with real-time voice in all of them.
24/7 availability with instant, consistent responses. No sick days, no timezones, no wait times.
RAG-powered answers from your documents, not generic internet content. Citations included.
SOC 2, FERPA, COPPA, WCAG AA compliant. AES-256 + TLS 1.3 encryption. 31 granular entitlements. Full audit trails.
***
## Administration & Analytics
Full control over Digital Twins, users, conversations, and analytics through the Administrative Panel.
Manage every Twin: AI model selection, tool permissions, knowledge bases, branding, and user access.
Enroll users, assign roles, manage permissions, and track engagement across your organization.
Review conversation transcripts, monitor quality, and gather insights from every interaction.
Control exactly what each user can do — toggle granular permissions across institutions, users, and histories to protect access to resources.
Track usage, credit consumption, model performance, and user engagement over time.
Customize your Twins with your institution's branding — logos, colors, and welcome screens. See the [Personalization guide](/mdx/admin-guide/personalization).
***
## How It All Fits Together
```mermaid theme={null}
flowchart TD
AC["Accounts & Billing"]
TO["39 Built-in Tools"]
GW["Google Workspace"]
US["Users & Roles"]
IN["Digital Twins"]
AC --> IN
TO --> IN
GW --> IN
US -->|Member| IN
US <-->|History| IN
IN --> AS["Assistants & Prompts"]
IN --> VA["Dynamic Variables"]
IN --> UP["IP Vault (RAG)"]
IN --> MC["MCP Servers"]
IN --> VC["Voice & Cloning"]
US <--> ME["Personal & Shared Memory"]
UP --> EM["Vector Embeddings"]
MC <-->|Twin-to-Twin| IN
```
***
## Create Your Digital Twin
Ready to build yours? Choose the path that fits your team:
Complete a quick form and our team handles everything — your Digital Twin will be ready in **24-48 hours**. Perfect for organizations that want expert setup.
Sign up, answer a few onboarding questions, and start using your Digital Twin in **10-20 minutes**. Full control from day one.
See the full step-by-step guide for all creation options →
***
## Explore the Platform
Learn how to communicate with your Digital Twin effectively.
Set up real-time voice with OpenAI, Gemini Live, or ElevenLabs voice cloning.
Build custom assistants to automate workflows and specialize your Twin.
Train your Digital Twin on your proprietary documents and content.
Connect Gmail, Drive, Calendar, Sheets, Classroom, and more.
Personal memory, shared memory, and conversation history explained.
Individual plans from free trials to enterprise tiers with volume savings.
LMS, Web SDK, MCP, Chat Completions, ElevenLabs, and REST APIs.