Platform

Everything you need to run AI in-house.

One platform for search, agents, research and finished deliverables — grounded in your knowledge, on any model, on infrastructure you control.

Any model — including your own

Route to Claude, GPT, Gemini, Llama or a private model you host. Switch providers without re-platforming, and keep sensitive workloads fully in-house.

Hybrid search + RAG

Vector and keyword retrieval over a Vespa index, reranked and grounded so answers cite the exact source passages — never invented.

Agents for every team

Package a prompt, knowledge and tools into a shareable agent. Finance, legal and support each get an assistant that knows their world.

Deep Research

Multi-step research that plans, retrieves across your sources, synthesises and writes a structured, cited report — in minutes, not days.

Finished deliverables

Not just a chat reply. Generate polished PowerPoint decks, formatted Word documents, charts and analyses ready to send.

On-brand by default

A Brand Kit applies your colours, fonts and logo to every deck and document, so output looks like it came from your team.

50+ connectors

Slack, Google Drive, SharePoint, Confluence, Jira, Salesforce and dozens more index in real time — with permissions synced from the source.

Permission-aware & cited

Every answer is scoped to what each user is allowed to see, and every claim links back to its source for full auditability.

How it works

Follow a request through the platform.

Scroll to watch one question move from a plain-language ask to finished, cited work.

01Ask in plain language
02Retrieve — permission-aware
03Reason with any model
04Deliver finished work
Ask in plain language
01

Ask in plain language

Someone asks a real question — no prompt engineering, no special syntax. Just what they need, in their own words.

02

Retrieve — permission-aware

Hadiq searches across every connected app, returning only what that person is allowed to see, with the exact source passages.

03

Reason with any model

Your chosen model — cloud or private — reasons over the retrieved knowledge, grounded so every claim traces back to a source.

04

Deliver finished work

Out comes a cited answer — or a polished, on-brand deck, document or analysis, ready to send.

How it works

One platform, end to end.

From a plain-language question to a finished, cited deliverable — grounded in your own knowledge, running on infrastructure you own.

Experience
Where you work
Chat · Admin console · Slack & Teams bot · MCP tools · REST API & SDK
Intelligence
The engine
Any modelHybrid search + RAGAgents & Deep ResearchDecks · Docs · ChartsCited & permission-aware
Connectors
50+ sources
SlackMicrosoft TeamsGoogle DriveSharePointOneDriveDropboxConfluenceJiraGitHubGitLabNotionSalesforce+ 50 more · permissions sync
Your infrastructure
You own it
PostgreSQL · Vespa · Redis & Celery — on-prem, private cloud or air-gapped · EU residency · open source
01Ask in plain language
02Retrieve — permission-aware
03Reason with any model
04Deliver — cited answer or finished doc

Deploy

Runs where you do.

Self-host in one command, scale on Kubernetes, or run fully air-gapped for the most sensitive environments.

Self-hosted (Docker)

One command to run the full stack on your own server.

Kubernetes / Helm

Production-grade, horizontally scalable deployments.

Air-gapped

Run with zero outbound connectivity for the most sensitive environments.

Get started

Own your intelligence.

Deploy Hadiq on your own infrastructure and give every team a permission-aware assistant that produces the work.