Upload and organize documents, ask source-backed questions, then use DeepSpace to research, draft, save notes, and create exportable deliverables. Add your preferred cloud or local AI provider, and connect supported apps only when you choose to authorize them.
Supported Sources & Connections
Start with supported files in Documents Hub, then connect a reviewed remote service only when it is useful to your work. Remote access is never implied: it depends on OAuth consent, connection health, selected scopes, and workspace policy.
GitHub
MCP · OAuth setup
Google Drive
MCP · developer preview
Google Calendar
MCP · developer preview
Google Chat
MCP · developer preview
Google People
MCP · developer preview
Web Search
Provider capability
Document processing
DOCX
Document processing
Gmail
MCP · developer preview
Document processing and web research | AverQel shows processing, extraction, retry, and quality states so users can see whether a source is ready instead of assuming every import succeeded.
Queued
Upload accepted | job created | worker dispatch
Prepare
Secure storage | source metadata | processing job
Parse
Supported extractor route | text coverage | quality signals
Chunk
Structured blocks | overlap windows | sanitized chunks
Embed
Batch vectors | provider metadata | chunk embeddings
Index
Chunk records | embeddings stored | query-ready state
How It Works
AverQel is not only a file bucket or a generic chatbot. It gives users a practical path: create a private workspace, organize material, ask grounded questions, do deeper work, and connect external apps only under explicit controls.
Start with a cloud or local AI provider, then upload supported documents into the Documents Hub. Connected apps are optional and are authorized separately through the MCP marketplace.
Documents Hub for supported files and source inspection
Optional supported app connections through OAuth
Provider, connection, and processing status stay visible
AverQel processes supported documents into searchable context. You can follow accepted, processing, indexed, retry, and extraction-quality states instead of assuming every file succeeded.
Parse | chunk | embed | index | source inspection
Grounded retrieval context for accessible documents
Retry and quarantine paths for problematic data
Use Grounded Query when evidence from documents is the priority. Use DeepSpace when you need research, drafting, notes, memory, or a controlled tool-assisted workflow in the same conversation.
Grounded answers, citations, and source inspection
DeepSpace activity, notes, memory, and saved history
External effects stay behind policy and approval controls
Organize documents into collections and share them through invitation and approval workflows. You choose exactly which documents to include, and owners keep control of what is visible.
Invite by collection code, approve or deny requests
Selective document inclusion, never forced global sharing
Owner and shared roles with distinct permissions
One Connected Workspace
Collections, Documents Hub, and DeepSpace meet in one permission-aware workspace. The interface below is an illustrated product flow based on the real surfaces, not a claim that every account is connected automatically.
A shared source, not a copied source.
Invite members, exchange encrypted messages and files, then add the exact documents the collection is allowed to use.
Room conversation
Encrypted messages · 3 members
The Real Product Walkthrough
AverQel is a connected workspace, not a generic chat box. Follow the steps below to see exactly where documents, grounded answers, DeepSpace, providers, and connected apps fit together.
01 · Providers
Add a cloud or local model provider for chat, research, embeddings, reranking, or web search. Your configured provider remains private to your account.
What you get
A ready AI runtime, chosen by you.
Runtime route
Cloud or local
Selected per workspace capability
Model inventory
Discovered
Context and capability metadata stay visible
Credentials
Protected
Masked in the user interface
Collections · Secure Collaboration
Collections let approved people work together without opening an entire workspace. Chat in real time, exchange encrypted attachments, and share the documents that belong to the project. Members see and use only the sources their collection permissions allow.
Shared documents remain in controlled storage. The collection grants access to the same source instead of creating a separate copy for every member.
E2EE
messages and media
Scoped
document access
Invite deliberately
Owners invite members and keep collection access explicit.
Talk in real time
Encrypted collection messages keep the project conversation together.
Send files securely
Share photos and file attachments as encrypted collection media.
Share the source once
Add an existing document to the collection instead of duplicating it for every member.
Document querying remains permission-aware: a member can use a shared source only when that collection and its document access are approved and available to the workspace.
Current Production Surfaces
Start with documents and evidence, then move into DeepSpace for deeper work. Collections, MCP connections, and provider control remain visible parts of the same workspace.
Document Surface
Bring supported files into one private workspace. Follow processing progress, inspect extracted text and chunks, review document state, and download the original file.
Retrieval Surface
Ask evidence-backed questions over the documents you can access. Results stay connected to source material, citations, and inspection flows.
Organization Surface
Create focused document sets for projects, teams, or topics. Collection ownership and sharing rules keep the scope deliberate rather than making all content globally visible.
Execution Surface
Use DeepSpace as a focused productivity chat for research, drafting, analysis, memory, and safe tool-assisted work.
Deliverable Surface
Move from chat into a real working surface with split layout, drafts, exports, and note-driven execution support.
Integration Surface
Authorize supported remote apps through OAuth. Each external tool is checked for ownership, connection health, catalog freshness, policy, and approval before use.
Runtime Surface
Choose the configured cloud or local runtime behind your work. Provider credentials are private to the account that adds them and are not exposed in the interface.
Runtime Commitments
Live from the current build
The homepage points users toward the actual shipped surfaces: grounded query, DeepSpace, the workspace editor, memory, connectors, provider control, and the security model that keeps them isolated.
Read the current product docsBeyond the Answer
AverQel is designed to help users continue work, not only generate a one-time reply. Explore the workspace capabilities below.
Workspace capability
DeepSpace can retain approved preferences and useful workspace facts across conversations. Memory is visible, editable, searchable, exportable, and removable by the user.
A Product Story, Not a Card Wall
This is an interactive product illustration, not a fabricated screenshot. It shows how the actual surfaces connect in a deliberate, user-controlled sequence.
Every stage has a visible boundary: sources, answer evidence, workspace work, then any authorized external service.
A workspace tree keeps files, folders, and shared sources visible at the start of the route.
A query trace shows the evidence path before a result becomes part of the work.
The active workspace turns that result into a draft, note, or next action with visible progress.
A final boundary makes external actions explicit: connection health, policy, and approval are visible.
Made for Real Work
AverQel adapts to the source material and the task. The workflow remains clear: organize context, ask better questions, make the result useful, and stay in control.
Upload source material, ask grounded questions, compare evidence, then turn the result into structured notes or an exportable draft.
Create focused collections, keep project context organized, use DeepSpace to draft and analyze, and save useful outcomes for the next session.
Use local or cloud providers, organize reference files, inspect supported code and text files in the Library, and connect approved tools when they are needed.
Turn a shared project into a clear handoff: discuss the work, keep decisions with the source material, and move the finished result into a usable deliverable.
Platform Features
Documents, grounded retrieval, DeepSpace, notes, MCP connections, providers, and security boundaries work together while remaining visible and user-controlled.
AverQel turns supported documents into inspectable, retrievable context. Processing state, extracted text, chunks, source details, retry paths, and quality signals remain visible.
DeepSpace is the focused workspace for research, drafting, analysis, notes, memory, and tool-assisted work. Users can follow visible activity, preserve useful work, and recover saved conversation history.
Move from an answer into editable notes, drafts, equations, diagrams, and exportable deliverables without leaving the workspace.
Connect supported remote MCP services through OAuth. Every external call remains subject to ownership, connection status, tool policy, risk limits, and approvals.
Use configured cloud providers or local runtimes without exposing secret values. Each provider belongs to the account that added it and can be selected for the relevant workflow.
AverQel keeps ownership, tenant isolation, encrypted provider and OAuth credentials, workspace policy, approval controls, and audit-safe execution boundaries in the product path.
Control Is Part of the Product
AverQel is built so users can see what is connected, choose what is allowed, and retain practical control over their account, data, and external actions.
MCP connection lifecycle
Connection health and tool availability are checked at run time. A connected account is not a promise that every remote provider will always be available.
Security & Trust
AverQel treats privacy, durability, and control as one runtime contract. DeepSpace keeps authoritative execution state in PostgreSQL, uses Redis only for live projections, and carries tenant isolation, encrypted secrets, workspace policy, approvals, and audit redaction through the execution path.
DeepSpace stores conversation messages and memory in PostgreSQL. Transient service state cannot authorize or replace tenant-scoped records.
Provider secrets and connector OAuth credentials remain encrypted, masked in responses, and protected by the existing provider and connector security boundaries.
Tenant and user ownership is carried through runs, nodes, events, approvals, checkpoints, leases, and tool records. Workspace policy and approval gates remain part of execution.
Persisted assistant messages let authorized users recover the visible thread after a browser or API interruption.
Built With
No experimental frameworks. Every component in the stack is well-documented, battle-tested, and actively maintained.
Start with a provider and your documents. Move from source-backed answers into DeepSpace work, saved notes, and exportable deliverables while connections remain under your control.