Capture
Upload files, media, or notes to your secure workspace.
How It Works
Your private AI assistant with powerful capabilities
100% Local AI
Encrypted Storage
Data Control
The Chat Experience
How it works
From case files to answers in minutes
A transparent workflow designed for clarity, compliance, and speed.
Secure
We encrypt and isolate everything before processing.
Analyze
Local AI runs retrieval, transcription, and reasoning.
Deliver
Receive auditable responses and export-ready reports.
Document Intelligence
Hybrid retrieval, not just search
Your question is first sharpened, then runs several searches at once — the results are fused, reranked, expanded to full sections, and can be verified before you read a word.
Upload Documents
Upload PDF, Word (.docx), Excel (.xlsx), or plain text documents to your secure AI Document Store. Files are processed locally on our servers — no external services involved.
Structure-Aware Processing
Your documents are split along their real structure — chapters, headings, pages and spreadsheet tabs — so every snippet keeps its context. A report section is never torn apart mid-thought.
Vector Embeddings
Every section is indexed twice: once by meaning (multilingual embeddings, 100+ languages) and once by exact words — so both “what it’s about” and “the exact invoice number” are findable.
Query Understanding
Before searching, a small local model rephrases your question and adds a keyword-focused variant — so vague or follow-up questions (“the second invoice”) still find the right passages.
Semantic search
Finds passages that mean the same thing as your question, even without matching words.
Keyword full-text
Catches exact terms, names, and codes that semantic search alone can miss.
Reciprocal-rank fusion
All result lists — every query variant, both methods — are merged into one ranked list, favoring passages that multiple searches agree on.
Relevance reranking
A dedicated reranking pass re-scores the merged candidates against your exact question.
Token-budgeted context
The most relevant, diverse passages fill a fixed context budget, capped per document so one file can't crowd out the rest.
Full-Section Context
The best-matching snippets are expanded to their surrounding section — the model reads the whole passage exactly as it appears in your document, not an isolated fragment.
Verified Answers
Optionally, every factual statement in the answer is checked against your sources after delivery — a green badge means each claim traces back to a document; anything unverified is flagged.
Up to +0
higher recall at 10,000-document scale
Internal benchmark, 10,000-document corpus.
Memory
How memory works
Dolos stores memory in an encrypted MemPalace, mines it in the background, and injects only bounded relevant context back into replies.
Memory flow
Chat turn
A completed user and assistant exchange becomes an immutable transcript shard.
Append shard
That shard is written into your encrypted memory workspace without delaying the response.
Sleep mine
When you go idle, background sleep mines the shard into exchange and general memories.
Palace drawers
Memories are organized into wings and rooms so you can browse, export, and delete them.
Wake-up + search
Before a reply, Dolos loads compact wake-up context and searches only the most relevant drawers.
Injected system message
Only a small bounded memory block is added to the system prompt, never your whole memory database.
Memory is not dumped into every prompt. Dolos adds a compact wake-up block plus a few relevant hits so responses stay fast and focused.
- Use the master Memory switch to pause both learning and memory-based personalization.
- Advanced controls let you separately manage background learning and chat-time memory usage.
- In Settings, you can review, export, delete drawers, remove rooms, or clear whole wings whenever you want.
- Turning memory off does not delete existing memories. It only pauses creation and injection until you turn it back on.
Your Memory Database stays fully manageable in Settings, including export, cleanup, and complete reset.
End-to-End Encryption
End-to-End Encrypted
100% end-to-end encrypted. Every message, file, AI response, API call, and real-time event is encrypted before it leaves your device — and server responses are encrypted before they leave ours. Not even our hosting provider can read your data.
100% Local AI
All AI inference runs on our servers - your data never leaves our infrastructure
Encrypted Storage
Your files are stored in AES-256 encrypted containers
Data Control at Your Fingertips
You choose when external APIs can access your data. Core AI runs locally; optional tools like search and translation use external services only when you invoke them.
Transport Layer (TLS)
All API and socket traffic is transmitted through TLS 1.3 to protect data on the wire and prevent interception attacks in transit.
End-to-End Payload Layer
A per-session AES-256-GCM key is negotiated at connection time via RSA-OAEP key exchange. Every outgoing request and every incoming response is encrypted with this session key. No plaintext ever crosses the network — not even through TLS-terminating proxies like Cloudflare.
Field-Level Database Layer
Sensitive DB fields are encrypted with AES-256-GCM before persistence, so database storage and snapshots do not contain plaintext values.
Communication Layer
Every real-time SocketIO event — chat messages, AI streaming responses, tool status updates, typing indicators, file processing notifications — is individually encrypted with AES-256-GCM session keys. Both directions: client-to-server and server-to-client.
File Layer
Every file interaction is encrypted end-to-end. Uploads are encrypted in your browser before sending. Downloads use per-file AES-256-GCM keys wrapped with your RSA public key. Document previews, thumbnails, and editor content are encrypted with your session key.
Sandboxed Execution Layer
Generated scripts and code-block runs execute inside an OS-level sandbox with restricted filesystem and network access, isolated from your main environment even when Agent Mode writes and runs code.
Your Data
E2EE Envelope
TLS Tunnel
Secure Processing
Field Encryption
Encrypted At Rest
What's Protected
- Chat messages and conversations
- Uploaded files and attachments
- Audio recordings and transcriptions
- Documents and RAG content
- Images, video files, and shared media
- Persisted DB fields and sensitive metadata
- All server responses (API data, chat history, file lists, settings)
- Real-time AI streaming, tool results, and status updates
Cryptographic Standard
Key Exchange
RSA-OAEP-256 (3072-bit)
E2EE Payload Encryption
AES-256-GCM
Transport Security
TLS 1.3
Database Field Encryption
AES-256-GCM (DB Fields)
At connection time, your browser and our server negotiate a session key: RSA-OAEP-256 (3072-bit) wraps an ephemeral AES-256-GCM key. All subsequent communication — requests, responses, SocketIO events, file transfers — uses this session key for sub-millisecond encryption. Database fields are independently encrypted with AES-256-GCM and PBKDF2 key derivation.
Verify Your Encryption
Click your username in the sidebar to view your E2EE status, key fingerprint, and active algorithms.
Powerful Tools
Powerful Tools
Each tool can run alone or be orchestrated by Agent Mode for multi-step execution.
Documents & Files
Document Search
100% local and secure. Your documents are processed entirely on our servers using advanced semantic search and intelligent reranking. No external services, no data leaves your environment. Search across all your uploaded documents with AI-powered understanding.
File Creation
Generate professional Word documents and Excel spreadsheets directly from your conversations. Export reports, chat protocols, research findings, transcripts, and more. Customize your name display for reports in Settings under 'Reporting Names'.
Build timelines & pivot tables
Generate .xlsx timelines pivoted by actor, location, or IOC.
Merge & deduplicate CSVs
Merge multiple watchlists, normalize columns, remove duplicates with fuzzy matching, and export a clean CSV ready for investigation workflows.
Generate DOCX/XLSX/CSV
Generate reports, spreadsheets, or CSV exports with sources.
Analyze email files
Import .eml files, extract text, metadata, and analyze with AI.
Media Analysis
Analyze any image with AI
Advanced local visual AI analyzes image content, identifies objects, describes scenes, and answers questions about images. Extracts structured data from documents, receipts, and forms — even in poor conditions beyond traditional OCR. Tables, text, and forms convert to clean markdown.
Transcript + scene highlights
Upload video files to generate full transcripts with time- and scene-markers, key events, and searchable segments for faster review and reporting.
Transcribe audio & video
A local transcription engine processes audio and video entirely on our servers — no external API calls. Supports 99 languages with automatic detection, and automatically extracts audio from video files.
Talk to your AI assistant
Enable voice responses from the AI. Choose between local TTS or a premium provider for natural-sounding voices. Configure your preferred voice and provider in Settings.
Translation & Language
Translate 400+ languages
On-device translation across 400+ language pairs — text, full documents, and OCR-extracted image text — with formatting preserved and files ready to export instantly.
Location & OSINT
Find where a photo was taken
Powered by Forensic Intelligence's GeoInspector API. Locate where a photo was taken using advanced topographical analysis and multiple specialized AI models. Get coordinates, location labels, scene classification, and heatmap visualizations — all integrated directly in chat.
Device Location
Access the user's current device coordinates via browser geolocation.
Spector OSINT Agent
Guided Spector investigations: manage cases, inspect entities, choose compatible providers, and run temporary or case-linked OSINT searches.
Memory
AI that remembers you
Your personal AI memory — securely stored on our servers. The AI remembers your preferences, biographical details, and instructions across conversations. View, edit, or delete your memories anytime in Settings. All data encrypted and never shared.
Search & Research
Live topics, current events
Anonymized web search for current events and live topics, returned with source summaries. Your identity is never exposed to the search provider. Configure up to 20 sources per search in Settings.
Source-linked answers
Multi-step, source-linked research that autonomously searches many sources, cross-checks findings, and synthesizes a cited report — anonymized, so your identity is never exposed to the research provider.
Crawl websites, import files, and analyze instantly
Built-in tool by
— crawl any website, directly import files from scraped pages into your secure cloud, and process everything with AI instantly.
Web scraping is not legal in every country, so you must check local laws before using this tool.
Development
Generate scripts
Built-in script generator by
— create forensic scripts with responsible defaults, no ethical pitfalls.
Red team scripts & tools
Create red team scripts with ease via
— run file manipulation tools in a secure sandbox environment.
Skill Agent
Advanced file generation agent for complex document creation tasks.
Agent Mode Orchestration
When extreme reasoning or external intelligence is needed, launch Agents with goals and multi-step plans. Agents can spawn focused sub-agents, and each sub-agent can use the same tools to complete the mission faster.
Planning Phase
Execution Phase
File Delivery Guard
Agent Synthesis
0+
Expert tools
0
Tool categories
0+
Export formats
0-bit
Encryption