PRODUCT · EXPLICIO PLATFORM · THE COGNITIVE FIRST LAYER

The assistant that asks questions before knowledge is lost – and feeds every system in the house.

Explicio sets in as the first layer where nothing is written down today: our active AI assistant collects experience knowledge from voice messages, videos, photos and chats, links it via connectors with your documents, maintenance reports and measurement series, and forms a knowledge graph in which every answer names its exact source.

Through our integrated MCP server and the agent API, both your employees and all higher-level systems, AI agents and third-party tools can build directly on this verified foundation.

Explicio has four parts: capture from around 20 channels, follow-up questions when something is missing, a knowledge graph built from the results, and a search that answers with a source reference. Limits and exclusions are stated on the same page as the capabilities.

Request trial accessPricing & Configurator▶ 90-second product walkthrough video
Channels & Connectors
Around 20 messenger channels + around 70 possible connectors, from M365, Drive and Confluence to maintenance reports & measurement series
Formats
Voice, video, photo, PDF up to 400 pages, email mailbox, ZIP archives & time series
Languages
37 languages live – the answer always comes in the language of the question
Compute location
EU cloud on hardware operated by Explicio in Germany or entirely local on the Explicio Box
Billing & API
You pay for the knowledge, not the chair – no seat licences, MCP/API included
HOW EXPLICIO WORKS AS THE FIRST LAYER

Capture · Connect · Synthesis · Graph

Four steps that happen on the side during everyday work. Nobody has to maintain a wiki – and every higher-level system benefits.

  1. CAPTURE01Hand over knowledge instead of writing it downA voice message on the way to the machine, a photo of the type plate, a video by smartphone or smart glasses – directly through the channel your people already use.
  2. CONNECT02Connect reports, repositories & measurement seriesThrough ready-made and custom connectors, SharePoint/M365, Confluence, email mailboxes, maintenance and service reports, and time series from measurements flow straight in.
  3. SYNTHESIS03Explicio extracts the essence and asks follow-up questionsProcesses, systems, rules and responsible persons are recognised and linked. If something is unclear or fits an existing node, the assistant asks a targeted question in the same channel.
  4. GRAPH & MCP04A navigator for people and AI agentsEvery piece of knowledge is tied to its source with page, timestamp or measured value. Whoever asks – whether an employee in the chat or a higher-level AI agent via the MCP API – gets the answer with the exact source reference.
PLATFORM DEEP DIVE · 90 SECONDS
ExplicioExplicioKNOWLEDGE IN · CHANNELS, FORMATS & CONNECTORS

Through the channel your team already uses – and directly from your systems.

No new tool, no training for the workforce: human experience knowledge comes in through around 20 familiar messenger channels, while our connectors automatically synchronise your file repositories, maintenance reports and measurement series.

KNOWLEDGE IN · FORMATS
FORMATS
Smartphone videoScreen recordingAction cam (GoPro)Smart glasses (Ray-Ban)Voice messagePhoto & type platePDF up to 400 pagesZIP archive & mailbox
All connectors at a glance →
WHERE CLASSIC TOOLS STOP

Explicio asks follow-up questions on its own.

Classic AI search only retrieves text where someone has already written something. Explicio actively elicits the knowledge that exists only in people's heads, keeps it up to date and ruthlessly uncovers contradictions between sources.

LINK“Do these belong together?”If two pieces of knowledge fit together but are not yet connected, the assistant asks exactly one short yes/no question. Answered confidently, they are linked; answered in the negative, it is learned – the same question never comes back.
CURRENCY“Is this still valid?”Knowledge gets a fixed expiry date. When it expires, the assistant automatically asks the contributor in their own channel. Outdated knowledge is clearly flagged in answers, not quietly hidden.
TRUSTProposed, confirmed, supersededKnowledge goes through a genuine trust status under the four-eyes principle. Whoever proposes may not confirm it themselves – and the review status is delivered with every answer.
CONTRADICTIONTwo statements, one transparent noticeIf two sources contradict each other (e.g. an old manual and a new maintenance report), the answer names both exact source references side by side – instead of silently deciding for one.
OUR OWN AI DEVELOPMENT · EXPLICIO PAIR4 & DETERMINISTIC KERNEL ARCHITECTURE96.4%Balanced hit rate on follow-up questions (550 test messages in 17 languages) · 95.5 % intent recognition in 19 languagesExplicio pair4: Our own decision model runs on the CPU – faster and more accurate than large language models.Before Explicio searches, our self-developed 307M-parameter model pair4 decides in just 62.8 ms directly on the CPU (without GPU load) whether a message is knowledge or a question, how a yes/no is meant, and whether the previous conversation is needed. On independent blind test sets, pair4 beats even large 35-billion-parameter models at intent recognition (95.5 % vs. 83.3 % for Qwen3.6-35B and 72.5 % for Nimble 9B). By separating the short decision call from the long extraction call, all structural graph decisions are also safeguarded against GPU kernel noise.Whitepaper: Decision model pair4 (PDF) Whitepaper: Reproducible AI decisions (PDF)
↓ PROCESSING KNOWLEDGE

Interactive knowledge graph — loads when it comes into view.

KNOWLEDGE OUT · FOR PEOPLE, AGENTS & THIRD-PARTY SYSTEMS

Find with the exact source reference – in the messenger or directly in your higher-level systems.

ANSWEREvery answer names its sourcesVoice message with timestamp, manual with page number, photo, chat or measurement series – per source, not as a blanket note at the end of the text. Anyone who wants to check an answer clicks straight through to the original evidence.
LANGUAGEIn the language of the question (37 languages)Anyone who asks in Polish gets the answer in Polish – even if the expert knowledge was originally recorded in German.
CHANNELBack to the same messengerThe answer arrives exactly where the question was asked. Live chat and search always take priority over any running bulk import.
VAULTAccess credentials without plaintext in the AI modelExplicio detects passwords and keys before every model call and stores them encrypted in the vault. The language model sees only placeholders; release follows the four-eyes principle.
OPEN INTERFACEMCP server and agent API (First Layer)Higher-level systems, your own AI agents and even specialised third-party tools dock on here seamlessly. They query through the same walls and with the same rights as a human – without a seat licence and at no extra charge.
EXISTING DATABulk import with trial runIngest entire ZIP archives, network drives and mailboxes in a controlled way. An upfront trial run estimates duration and scope, and the import runs throttled in the background without slowing down live chat.
RELIABLE SYSTEM BOUNDARIES · WORKS-COUNCIL- & AUDIT-PROOF

Clear boundaries for maximum trust in operations.

A tool that structures and secures knowledge – not one that evaluates people or autonomously controls plant and equipment.

PEOPLEExclusion of employee evaluationsThe assistant answers no questions whatsoever about a person's suitability, performance or reliability (actively detected and declined by pair4). It only shows responsibilities – there is no ranking anywhere.For works councils & data protection →
ACTIONNo controlling intervention in your systemsExplicio reads and delivers verified knowledge. It controls no machines, releases no payments and takes no personnel measures – every action with consequences remains with the human.
MODELSZero training on your customer dataYour data never flows into public models and never improves other tenants' models. Processing runs in isolation in our EU cloud or on your Explicio Box.
FORMATSHonest knowledge gaps instead of hallucinationPDF, text, image, video, audio and measurement series are understood; special formats such as raw CAD files are stored as a protected attachment. Where knowledge is missing, the assistant names the gap instead of guessing.
TECHNICAL DIAGNOSIS · DEMO WITHOUT A CALL

First bring in knowledge. Then decide.

After a short conversation we set up your trial space. Then contribute a voice message or a file — and see where the first knowledge island lies. The 30 days only start after that.

Request a quoteView Explicio BoxView Explicio Grid
Duration
30 days from the first knowledge
Cost
€0
Licence
none — you pay for knowledge, not for seats
Data location
EU cloud on hardware operated by Explicio in Germany; box on request
Service statusTo the live status page →