The Complete AI System.
Governed. Traceable. Auditable.

iKB handles everything — retrieval, models, channels, storage, moderation, and audit. Integrate any system through our API for reliable, governed answers. Connect to virtually any storage platform. Run it on your infrastructure or ours.

Complete System Retrieval, models, channels, moderation, governance, and audit — all in one platform. Nothing to stitch together.
API-First Integration Connect your systems to iKB's API for reliable, governed AI answers. Build products and solutions on top.
Universal Storage 40+ cloud storage providers, databases, web crawlers, 25+ document formats. Connect to virtually any data source.
Award-winning
6 channels + REST API
Self-hosted or cloud
40+ storage connectors
Independent Evaluation

Tested Against the Best. Perfect Score.

20-question adversarial retrieval accuracy test across two dense regulatory documents. Evaluated by Claude Opus 4.6. iKB achieved a perfect 10.0/10 without thinking mode — outperforming both competitors that had thinking enabled.

iKB no thinking
10.0
Gemini 3.1 thinking
9.3
ChatGPT 5.4 thinking
8.8
0
Hallucinations
0
Retrieval Errors
20/20
Perfect Answers
437
Pages Tested

EPSU GDPR Briefing

40 pages • Narrative regulatory content • 10 questions

iKB
10.0
Gemini
9.1
ChatGPT
8.4

PCI DSS v4.0.1

397 pages • Dense three-column tables • 10 questions

iKB
10.0
Gemini
9.5
ChatGPT
9.2

Where Competitors Failed

ChatGPT 5.4
Retrieved wrong requirement (PCI Q5) — returned exclusions from 8.3.10.1 instead of 8.3.9. Flattened uncertainty (GDPR Q1, Q4, Q9) — turned hedged language into definitive claims. Score: 3/10 on worst question.
Gemini 3.1
Hallucinated an exemption (PCI Q6) — fabricated an MFA exemption category not in the standard. Cross-contaminated chunks (GDPR Q1) — fused page 29 content into page 7 argument. Score: 5/10 on worst question.
iKB
Zero hallucinations. Zero retrieval errors. Perfect table-row indexing on 397-page dense tabular document. Preserved source uncertainty. No thinking mode required.

Evaluated by Claude Opus 4.6 (Anthropic) • 28 March 2026 • 5-dimension scoring: Accuracy, Completeness, Hallucination, Confidence Calibration, Exception Awareness

Infrastructure Guarantees

Every AI Interaction — Governed, Traced, Auditable

Accuracy

Every response grounded in source data with citations. Quality scoring on faithfulness, relevance, and groundedness. Automatic regeneration on low scores — across every channel your solution uses.

Governance

Four-pillar immutable audit system built into the pipeline, not bolted on. Audit logs, prompt versioning, config tracking, and moderation events — enforced identically across all 6 channels and the REST API.

Traceability

Complete chain of custody for every interaction through the pipeline. Prompt changes versioned, 47+ config fields tracked with before/after snapshots. Every action captured with actor, IP, and timestamp — regardless of which channel triggered it.

Governance Layer

Governance Built Into the Pipeline

Four independent, immutable logging systems that every channel inherits automatically. Consent management, content moderation, and DSAR compliance — enforced at the infrastructure level.

4
Audit Pillars

Logs, Prompt Versions, Config Changes, Moderation Events — all append-only, never deletable

47+
Tracked Fields

Before/after snapshots for every topic setting change with IP attribution

13
Moderation Categories

Fail-closed default, per-topic, per-channel, every call logged with response time

0
Deletable Records

Immutable audit trail. Deletion returns 403 Forbidden. Full non-repudiation.

Content Moderation

Built-in 13-category moderation using OpenAI's omni-moderation model, free of charge.

Fail-Closed Default Per-Topic Toggle Per-Channel Coverage Performance Tracking Translated Warnings

GDPR & Data Privacy

Dedicated DSAR module gives superadmins complete GDPR compliance tooling.

Search by Email Export (Art. 15) Purge (Art. 17) Audit-Logged Compliance-Safe

Complete Sovereignty — Run Fully Offline

iKB is built to operate in air-gapped, zero-trust environments with absolutely no external dependencies. Deploy on your own GPU infrastructure using open source AI models through vLLM, Ollama, or any OpenAI-compatible endpoint. Every component — inference, embeddings, moderation, storage — runs locally. No data ever leaves your network. No cloud calls. No phone-home. Full operational sovereignty for defence, government, finance, and any environment where data must never cross the perimeter.

vLLM Ollama OpenAI-Compatible Local GPU Inference Air-Gapped Deployment Zero External Dependencies On-Premise Embeddings No Cloud Calls
Immutability Append-only records
Fail-Safe Logging never breaks ops
Privacy by Design Auto-redaction
Non-Repudiation IP + actor + timestamp
Exports CSV / ZIP up to 10K records
RAG Pipeline

One Pipeline. Every Channel. Every Model.

Unified 8-stage retrieval pipeline shared across all channels. Plug in any LLM via LiteLLM, connect any data source, and get quality scoring, per-user memory, and query decomposition automatically.

iKB Memory

Per-user semantic memory via Mem0 + pgvector. AI remembers preferences and facts across sessions with privacy-first incognito mode.

Unified RAG Pipeline

8-stage pipeline: HyDE, query decomposition, reranking, neighbourhood expansion — shared across all 6 channels.

GraphRAG

Cross-document entity relationships, hybrid graph+vector queries, 15+ configurable settings.

Multi-Model via LiteLLM OpenAI, Anthropic, Google, Cohere, Cerebras, LocalAI. Unified gateway with cost tracking.
Source Citations Every answer shows source docs with page numbers and direct links.
Multi-Language Ask in any language, answer in any language. Full i18n support.
Spreadsheet Analytics Natural language queries with sandboxed Python, multi-sheet JOINs.
RAG Strategy Recommender Analyzes content and recommends: vector, hybrid, or GraphRAG.
18 Output Skills Built-in skills for summaries, reports, comparisons, checklists, and Word/Excel/PPT export.
Knowledge Organization Separate knowledge bases per product/team with independent AI settings.
Per-Channel Instructions Different master AI instructions per channel. 7 rendering engines including inline SVG graphics.
Follow-Up Chips & Skeleton Loading AI-generated follow-up suggestions after each response. Streaming skeleton loader for smoother UX.
Security

Enterprise Security

147-finding security audit. Egress policy system, DACL permissions, nonce-based tool delimiters, Fail2Ban, and 2,251 automated tests.

Dual-Key Encryption

AES-256-GCM with user key + admin key. All secrets Fernet-encrypted. Sensitive values auto-redacted in logs.

Enterprise SSO

Authentik OIDC with headless auth, JIT provisioning, group sync. Three modes: SSO-only, hybrid, local-only.

App Hardening

SSRF blocking, ODBC injection prevention, CSP nonces, CSRF, HSTS, XSS remediation, Fail2Ban integration.

DACL Permissions

18 granular permission columns per admin. Deny-by-default, hierarchy enforcement, per-channel limits. Full audit logging.

Session Security

HTTPOnly cookies, strict SameSite, UUID v4 validation, TOTP 2FA, HMAC webhooks, IP whitelisting.

No External Training

Your documents are NEVER used to train AI models. Complete data sovereignty guaranteed.

Channel Layer

Plug In Any Channel. Same Pipeline. Same Governance.

6 channels connect to the same RAG pipeline with the same governance, traceability, and quality guarantees. Build customer-facing products, internal tools, or API integrations — all governed identically.

Native Chat
Web Widget
Teams
Slack
Chatwoot/OMNI
REST API

Web Chat Widget

Deploy on any website with a single script tag.

Voice STT/TTS Starter Forms Human Escalation AI Disclosure Lead Capture 3 Display Modes Light/Dark/Auto Domain Whitelist

Human Escalation

When AI can't help, users request human assistance directly from chat.

Per-Topic Toggle Email + Transcript Custom Labels Rate Limited Cross-Channel

External API Tools

Connect AI to live data and business systems with full governance.

MCP Connections REST Tools Parallel Execution Encrypted Auth Block Mode Execution Audit

SQL Functions & APIs

Schema-level discovery, admin annotations, query playground.

Auto-Discovery REST API Cloud Sync (40+) Chat Export Feedback System Governance Exports
Deploy & Scale

Flexible Deployment, Any Scale

Cloud SaaS

Fully managed, automatic updates and scaling

Private Cloud

Dedicated instance in your cloud region

On-Premise

Your infrastructure, air-gapped available

25+ Document Formats

PDFWordExcelCSVMarkdownRTFEPUBBibTeXDWGDXFEMLMSGPNGJPGGIFTIFFWEBPWeb Crawl

Magic byte validation, ClamAV scanning, ZIP bomb protection, UUID filename sanitization.

Performance

<500msQuery p95
100+Pages/min ingestion
10K+Concurrent users
99.9%Uptime SLA

The iKB Difference

Piecing It Together Yourself
iKB — Complete System
Stitch together retrieval, models, channels, audit
Everything in one governed platform
Each integration is a custom project
One API for reliable, governed answers
Connect storage sources one at a time
40+ storage providers + 25+ document formats
Governance bolted on later (or never)
Governance, traceability, and audit from day one
Locked to one model or one cloud
Any model, any cloud, or fully local
Build On iKB

What Teams Build on iKB Infrastructure

Governed AI Products

Build customer-facing AI products with immutable audit trails, prompt versioning, and content moderation baked into the infrastructure. Ship governed AI without building governance.

Sovereign AI Infrastructure

Deploy on your own infrastructure with complete data sovereignty. Air-gapped, local models via vLLM/Ollama, SSO integration. Full traceability for public accountability.

Multi-Channel AI Solutions

Plug Slack, Teams, WhatsApp, web widget, and REST API into one governed pipeline. Each channel gets the same retrieval quality, moderation, and audit trail automatically.

Internal AI Platforms

Build internal knowledge platforms with 25+ document formats, per-team knowledge bases, skills system, and governance exports. Teams get consistent, traceable answers from your data.

Deploy across 6 channels: Native Chat, Web Widget, Teams, Slack, Chatwoot/OMNI, REST API. Learn more →

One system. Everything you need.

Self-hosted or cloud. Any model. Any storage. Complete governance.