Menu
Home
Use Cases
Events
Open Platform

KriftAI Product

AI Knowledge Workspace with Governed Multi-Persona Intelligence

AI-powered knowledge workspace with governed chat, artifacts, voice response, and API access

20+

AI Agents

3

Response Modes

API

External Access

100%

Governed

Overview

Your foundational AI copilot

The KriftAI Vanguard is a governed AI knowledge workspace for professional and enterprise teams — combining multi-persona intelligence, semantic knowledge retrieval, and artifact-grounded reasoning into a single deployable environment. Use it internally for research, analysis, and document drafting, or deploy it as an AI backend for any external application via API.

Unlike specialized workspaces with domain-specific workflows and stages, the KriftAI Vanguard operates as a flexible, all-purpose AI agent that adapts to any subject matter — and serves as the foundation for deploying governed AI agents powered by custom personas to external applications via API.

Core Capabilities

Everything you need in one workspace

Conversational AI Chat

Multi-turn conversations with full context retention across messages
Markdown rendering — headers, tables, code blocks, bold, links
20+ built-in agent personas or create custom personas with 10,000+ character definitions
Switch agent personas mid-project — each persona brings its own expertise and boundaries
Real-time web search for current events and external research
Conversation history — navigate between conversations within a project

Artifact Management (Knowledge Base)

Upload PDF, Word, text, CSV, JSON, images via drag-and-drop
Documents auto-chunked, embedded, and retrieved by semantic relevance
AI references uploaded documents using artifact citation format
Three scoping levels: Personal, Organization, Project — controls visibility
Both direct and semantic similarity retrieval on every AI interaction
Unlimited document uploads — build comprehensive knowledge bases

Voice & Video Response Modes

Text mode (default) — standard markdown responses
Voice mode — each response gets a play button for audio playback (6 voice options)
Video mode — AI avatar speaks the response
Toggle per project via Integration Admin settings
Markdown stripped for clean speech output
Ideal for accessibility, voice advisors, and customer-facing applications

Per-Project API Keys

Generate API keys scoped to individual projects
Keys stored as SHA-256 hashes — raw key shown once, never stored in plaintext
Each key tied to one project — can only access that project's data and personas
Usage tracking with last-used date and total call count
One-click revocation — key immediately stops working
External apps use x-api-key header to call KriftAI's APIs

External Integration

Your AI backend for any application

The KriftAI Vanguard isn't just an internal tool — it's an AI backend. Create a persona, upload a knowledge base, generate an API key, and any external application can embed governed, artifact-aware AI responses.

01

Create a Persona

Define or select a persona with domain expertise, tone, and governance boundaries. The persona becomes the intelligence layer for your external app.

02

Upload Knowledge Base

Upload documents — FAQs, product docs, policies, procedures. The AI grounds every response in your actual content, not generic training data.

03

Generate API Key

Generate a project-scoped API key. External apps authenticate with x-api-key header and get governed, knowledge-grounded responses.

04

Deploy Anywhere

Customer support chatbot, product advisor, technical assistant, domain expert — any SaaS app can embed KriftAI personas with full governance and audit trail.

What the external app gets

Persona intelligence — domain expertise defined by you
Knowledge base grounding — from uploaded artifacts
Governance protection — input/output validation, audit trail
Voice output — returns audio URL alongside text (if enabled)
Stateless per request — no data retained by KriftAI

Governance Layer

Every interaction is governed

Governance is infrastructure-level, not prompt-level. Even if a persona's system prompt is ignored by the LLM, the output validator catches prohibited content before the user sees it.

Input Validation

Scans user input for prompt injection attempts (7 attack patterns) before reaching the LLM.

Output Validation

Scans AI response for prohibited content — credential leaks, jailbreak leaks, system prompt exposure — before returning to user.

Scope Enforcement

Personas configured with hard boundaries — code-level enforcement ensures the AI cannot break its core rules regardless of how users phrase requests.

Audit Logging

Every AI call logged: user, persona, input summary, output summary, model, governance results. Fully reconstructable decision trail.

Escalation (Break Glass)

If governance blocks a response, user can request override with documented reason. Admin reviews the escalation before release.

Credential Redaction

API keys, passwords, and credentials are automatically stripped from AI output if detected — preventing accidental exposure.

Your AI, Deployed Anywhere

Create a persona. Define its world. Deploy it anywhere.

The KriftAI Vanguard lets you create a custom AI persona — define its expertise, its tone, its boundaries, and its knowledge base — then deploy it as a governed AI agent inside your own website, blog, SaaS app, or customer portal. No AI expertise required.

01

Define the Persona

Set the name, expertise, tone of voice, and hard boundaries. A 10,000+ character definition that encodes exactly how this AI should behave — what it can discuss, what it refuses, how it responds.

02

Upload Its Knowledge

Upload the documents, FAQs, product guides, or policies that the persona should reference. The AI grounds every response in your actual content — not generic training data.

03

Set Governance Rules

Define what the persona cannot do. Block topics, prevent hallucination outside its knowledge base, enforce compliance rules. Code-level enforcement — not prompt suggestions.

04

Deploy with One API Key

Generate a project-scoped API key and embed the persona in your website, blog, customer portal, or SaaS application. Your visitors interact with your AI — governed, auditable, on-brand.

Think of it as creating your own AI employee — with defined expertise, clear boundaries, and a governed knowledge base — deployable anywhere with a single API call.

Use Cases

Built for any knowledge work

Research & Analysis

Upload documents, ask questions, get artifact-grounded analysis. Decision support with SWOT, cost-benefit, and options analysis across any domain.

Embedded AI Chatbot

Create a persona, upload FAQs, generate an API key — deploy a governed customer support chatbot or product advisor in your own SaaS application.

Voice-Enabled Advisor

Professional advisor that speaks responses. Financial, legal, medical consultation — or a customer-facing voice assistant embedded in a website.

Document Drafting

Reports, proposals, emails, documentation — drafted with context from your uploaded knowledge base and refined through multi-turn conversation.

Multi-Persona Research

Switch between specialist personas within one project. Get perspectives from different experts on the same question using the same knowledge base.

Knowledge Synthesis

Upload documents from multiple sources. The AI synthesizes across them — finding connections, contradictions, and insights humans might miss.

Standard vs Specialized

Choose the right workspace

KriftAI Vanguard
Specialized Workspaces
Generalist or custom persona
Domain expert persona (BA, SEO, GMP, Legal, etc.)
No workflow stages — flexible chat
4–9 structured stages with gates and approvals
No domain-specific tools
Domain-specific analysis tools built in
Chat interface only
Stage drawer, workflow panel, guided process
Platform-level governance (universal rules)
Platform + workspace-specific governance rules
General knowledge work + external API integration
Domain-specific professional workflows

Start with the KriftAI Vanguard

The foundation of every KriftAI deployment. Whether you're using it for internal knowledge work or embedding governed AI into your own application — it starts here.