Product

MOST Insight: the AI layer on top of all your documents

MOST LLC is an artificial-intelligence company. Our product, MOST Insight, applies retrieval-augmented generation to your organisation’s own files so that every employee can get a trustworthy, cited answer in seconds — on the desktop, on the phone, in Teams or through an API.

AI at the core

Not a search box — an analyst that has read everything

Classic enterprise search returns a list of files. MOST Insight reads the relevant passages, reconciles them and writes the answer, with the evidence attached. Under the hood: dense embeddings, keyword retrieval, source tiering, query rewriting and large-context language models.

Ask, don’t search

Type a plain-language question. MOST Insight reads the relevant pages across every connected source and writes a concise answer with the exact documents it relied on.

Every answer is cited

Each response links to the source documents and excerpts behind it. The AI is instructed to never invent facts — if the answer is not in your documents, it says so.

Permission-aware by design

Access rights from Microsoft Entra ID, Google Workspace and your file systems are indexed with every document. Users only ever see answers built from files they are allowed to open.

Hybrid retrieval engine

Dense vector embeddings and BM25 keyword search are fused for high recall on both conversational questions and exact part numbers, invoice IDs or clause references.

Source tiering

Rank official policies above e-mail threads. Tiered sources let the assistant prioritise authoritative documents and flag contradictions instead of guessing.

Multilingual

Ask in one language, get answers from documents written in another. Built for international organisations with mixed-language archives.

Admin analytics

A pseudonymised question store shows what people ask, where the gaps in your documentation are and what each answer costs — without exposing individual identities.

Open integration API

Expose the assistant to other tools through a REST API and the Model Context Protocol (MCP), so your own agents and workflows can query company knowledge.

Chat where you work

Native web app, Microsoft Teams and Telegram bots, plus desktop and mobile apps. The same permission checks apply on every channel.

Everywhere you work

Search your desktop, your phone and every cloud

The same assistant, the same permission checks, on every device and channel.

Desktop apps

Windows, macOS and Linux agents index local folders and give instant answers from a hotkey or the menu bar.

Mobile apps

iOS and Android apps with secure sign-in; snap a photo of a document and ask about it.

Chat bots

Microsoft Teams, Slack and Telegram bots bound to the corporate identity of each user.

Web app

A fast, accessible web assistant with source previews, history and feedback.

API & MCP

REST endpoints and a Model Context Protocol server so your own agents can use company knowledge.

Admin console

Usage analytics, cost per answer, documentation gaps and connector health in one place.

How it works

Architecture

A containerised platform you can run in our cloud, your cloud or your own data centre.

MOST Insight architecture: sources, indexer, vector store, AI engine and clientsSOURCESMicrosoft 365Google WorkspaceDropbox · Box · S3Desktops & phonesNetwork sharesIdentity (Entra, Google)Indexercrawl · extract · OCRchunk · embedcapture access rightshigh-performance Rust serviceVector storedense embeddingsBM25 sparse indexper-chunk ACLsprivate per tenantAI enginehybrid retrievalpermission filterLLM answer + citationsNode.js / TypeScriptLanguage modelsAzure OpenAI · privateor on-premise GPUCLIENTSWeb appDesktop appsiOS & AndroidTeams · Slack · TelegramREST & MCP APIAdmin consolevectors + ACLcited answers · only from files the user may opencontinuous sync · nightly full pass
01

Connect your sources

Authorise Microsoft 365, Google Workspace, Dropbox, Box or any other connector in minutes. Install the lightweight desktop and mobile agents for local files.

02

Index & embed

Our high-performance indexer extracts text from Word, PDF, PowerPoint, Excel, e-mail and more, chunks it intelligently and stores embeddings together with the access rights of each file.

03

Ask questions

Users ask in the web app, Teams, Telegram or on their phone. The engine retrieves the best passages they are allowed to see and generates a cited answer.

04

Stay in sync

Changes are picked up continuously; a full consistency pass runs nightly. Permissions are refreshed on a schedule so revoked access disappears from answers automatically.

Formats

Reads what your people actually produce

Text is extracted natively from office formats and e-mail; scanned material goes through OCR.

  • Word (.docx, .doc)
  • PDF (digital & scanned with OCR)
  • PowerPoint (.pptx, .ppt)
  • Excel (.xlsx, .csv)
  • E-mail (.eml, .msg)
  • Google Docs / Sheets / Slides
  • Plain text & Markdown
  • HTML & web pages
  • Images of documents (OCR)

Ready to give your team instant answers?

Start on any plan today, or talk to us about an on-premise deployment for your organisation.