StellarBase
StellarBase Core Product

Your company's knowledge,
unified with AI.

Plug in your docs, emails, chats, databases, and images. StellarBase turns it all into a searchable knowledge base you can chat with, automate, and trust.

Self-hosted
or
EU EU cloud
·
Your data stays yours
At a glance

Everything you need in one platform

Six pillars that work together. Deploy once, scale endlessly.

Data Sources

50+ native integrations

Connect StellarBase to the tools your team already uses. Data flows in automatically via webhooks or scheduled syncs. Everything stays in sync — no manual imports, no data silos.

Communication
Slack Slack
Microsoft Teams Microsoft Teams
Q3 2026
Gmail Gmail
Google Chat Google Chat
+ more
Cloud Storage
Google Drive Google Drive
AWS S3 AWS S3
GCP Buckets GCP Buckets
Q4 2026
OneDrive OneDrive
iCloud iCloud
Azure Blob Azure Blob
Local Disk
+ Any S3-compatible
Productivity
Q3 2026
Notion Notion
Obsidian Obsidian
Google Docs Google Docs
Google Sheets Google Sheets
Google Slides Google Slides
Google Calendar Google Calendar
+ more
Microsoft 365
Word Word
Excel Excel
Q3 2026
PowerPoint PowerPoint
Outlook
+ more
SQL Databases
Q3 2026
PostgreSQL PostgreSQL
MySQL
SQLite
MS SQL Server
MariaDB
+ Any SQL source
NoSQL & More
Q4 2026
MongoDB
Elasticsearch
Redis
DynamoDB
Firestore
+ more

Any S3-compatible

MinIO, Backblaze, Wasabi, DO Spaces

REST API

Push data from any custom source

Webhooks & Feeds

Real-time from any system

File Formats

40+ formats, processed automatically

Documents, spreadsheets, presentations, images, audio, video, email, code, archives — everything your team produces is processed, indexed, and made searchable.

Documents

PDF
OCR + text extraction Live
DOCX / DOC
Full text + formatting Live
TXT / Markdown
Plain text Live
HTML
Content + link following Live
RTF / EPUB
Rich formats, e-books Q3 2026

Spreadsheets

XLSX / XLS
Excel with formulas Live
CSV / TSV
Delimited data Live
ODS / Parquet
Open + columnar Q3 2026

Presentations

PPTX / PPT
Slides + embedded media Live
ODP / Keynote
Open + Apple presentations Q4 2026

Images

PNG / JPG / WEBP
OCR + object recognition Live
SVG
Vector content extraction Live
TIFF / BMP / HEIC
Scanned + Apple photos Q3 2026

Audio

Soon
MP3 / WAV / M4A
Transcription + speaker detection Soon
FLAC / OGG / OPUS
Lossless + voice Soon

Video

Soon
MP4 / MOV / WEBM
Transcription + frame extraction Soon
AVI / MKV
Legacy formats Soon

Email

Soon
EML / MSG
Headers + body + attachments Soon
MBOX / PST
Mailbox archives Soon

Code & Data

Source Code
30+ languages, syntax-aware Live
JSON / XML / YAML
Structured data Live
SQL dumps
Schema + data Q3 2026

Archives & Chat

Soon
ZIP / RAR / 7Z
Recursive extraction Soon
Slack / Teams exports
Chat history Soon
Data Processing

Build your base — your data, your rules.

Every file you connect is processed locally. Sophisticated OCR, table extraction, image understanding, audio transcription — all handled by StellarBase engines. Your processed data stays on your disk. No vendor lock-in. Ever.

Processing Pipeline Customizable
Live
Ingestion Connect source, fetch content
Extraction Parse, OCR, transcribe
Enrichment Detect entities, relationships
Indexing Store in semantic graph
Add your own steps. Modify any stage. Plug in custom extractors.
Sophisticated extraction

Not just text. Everything.

Advanced OCR

Scanned documents, photos of receipts, handwritten notes, low-quality images. Multi-language, layout-aware, preserves reading order.

Printed textHandwriting100+ languagesMulti-columnFormsReceipts

Complex tables & structures

Merged cells, nested headers, multi-page tables, hierarchical structures. Preserves relationships between cells, columns, and rows.

Merged cellsNested headersMulti-pageFormulasPivot tablesCharts

Image understanding

Beyond OCR — scene understanding, object detection, chart interpretation, diagram parsing. Images become searchable content.

Object detectionChart parsingDiagramsFacesLogosScene context

Audio & video

Soon

Full transcription with speaker diarization, timestamps, sentiment. Video adds frame extraction, scene detection, on-screen text OCR.

TranscriptionSpeaker IDTimestampsFrame extractScene detectSentiment
Multimodal by architecture

Not a bolted-on embedding

Text, images, audio, video, tables, and structured data are first-class citizens at every layer. No dependency on external multimodal embeddings. No vendor lock-in to a model provider.

Text
Images
Audio
Video
Tables
Structured
Unified into one graph
One knowledge model. No separate silos. No re-embedding.

Everything you expect, and more

Complex relationships

StellarBase discovers how documents, people, concepts, and media connect. Revenue → EMEA → contracts → specific employees — all surfaced automatically.

Lightning-fast querying

Sub-5ms median latency. Indexed and optimized for real-time interactive use. Ask complex questions across millions of documents, get answers instantly.

Multimodal queries

Search by text, image, or both. "Find contracts similar to this screenshot." "Show charts that discuss revenue." Results span any modality.

Customizable pipeline

Add custom extractors. Run domain-specific enrichment. Skip stages you don't need. The pipeline is yours to shape.

Your data. Always.

Processed data lives on your disk, in your database. Export anytime. Move anywhere. Zero vendor lock-in — if we fail you, take everything and leave.

Full Transparency

Every step is yours to see,
inspect, and control.

No black boxes. No hidden processing. You see exactly what StellarBase extracted from every file, every step of the way. Review, correct, and approve what enters your knowledge base.

Inspectable at every stage

OCR output, table extraction, entity detection — all auditable.

Human-in-the-loop review

Flag uncertain extractions for manual approval before they enter the base.

Correct once, learn forever

Your corrections feed back into the pipeline. Accuracy improves with every review.

Full audit trail

Who processed what, when, with what confidence. Every action logged.

Extraction Review
invoice-q4.pdf
Vendor
98% confidence
Acme Corporation
Total Amount
99% confidence
€12,480.00
Due Date
72% — needs review
2026-03-15
Audit Trail
OCR complete 2.4s · v1.2
Table extracted 0.8s · 3 rows
Entities detected 0.3s · 7 entities
Indexed in graph Live ✓
AI Agents

Like hiring a new colleague.
Just faster.

Anyone can create a specialized agent. Connect the sources it should read. Define how it should behave. Give it a goal — it works until the goal is done. It remembers, learns, and gets better over time. Just like a real teammate.

Create Agent
01
Identity
Name
Financial Analyst
Type
Analyst
Instructions
"You analyze quarterly reports..."
02
Connect sources
Financial Reports
Bloomberg API
Internal Wiki
+ Add source
03
Tools & abilities
Search KB
Web access
Run workflow
Send email
Call API
Cite sources
Planning mode
ReAct + Reflection
Memory
Persistent, learns from feedback
Agent types

Choose your specialist

Researcher

Deep research across sources and the web, finds and synthesizes information

Analyst

Analyzes documents, extracts patterns, generates structured insights

Writer

Drafts reports, emails, documentation grounded in your data

Coordinator

Orchestrates other agents, delegates subtasks, aggregates results

Reviewer

Reviews outputs for quality, flags issues, approves or rejects

Operator

Executes external actions — API calls, data mutations, integrations

Goal-oriented execution

Works until the job is done

Give the agent a goal. It plans, acts, observes results, reflects, and iterates — automatically. Long-running tasks like "analyze these 2,000 contracts" just work.

Goal: "Analyze all Q4 contracts for liability risk"
Iter. 47/∞
Plan Break goal into steps
Act Execute next step
Observe Check results
Reflect Adjust approach
Repeats until goal is complete
Zero-Trust access — chained

Agents see only what you see

Every agent inherits the permissions of the user who invoked it. If you can't access a source, neither can the agent — even if the agent was configured with broader access. Permissions chain down, always.

Alice

Finance team

✓ Financial Reports ✗ HR Records
invokes
Financial Analyst

Agent config: all sources

Inherits Alice's permissions
reads
Financial Reports HR Records (denied)
Only sources Alice can read
Agents collaborate

Build specialized teams

Agents pass tasks to other agents, delegate subtasks, and share context. A coordinator breaks down a goal, specialists execute, a reviewer verifies — all transparent, all auditable, all under your supervision.

Workflow: Competitor Analysis Report In progress
Research Agent Gathers sources
Working...
Analyst Agent Extracts insights
Waiting
Writer Agent Drafts content
Waiting
Reviewer Agent Verifies output
Waiting
Every delegation is visible. Approve or intervene at any step.

Everything an agent can do

Grounded in your data

Every answer is based on your connected sources. Agents don't hallucinate — they cite what they know and say when they don't know.

Verified citations

Every claim links to the source document with page numbers and confidence scores. Click any citation to see the original. Full traceability, always.

Memory & learning

Persistent memory across sessions. Remembers context, learns from your corrections, adapts to your preferences. Gets better with every interaction.

Internet & deep research

Web search, content extraction, multi-source synthesis. Agents can research across the open web in addition to your private knowledge base.

External actions

Call any API, send emails, update databases, trigger integrations. Agents execute — not just talk. Every action logged and revocable.

Self-built workflows

For massive tasks, agents can design and execute their own workflows — parallelize work, batch processing, multi-step automation. You approve, they build.

Workflows Q3 2026

Deterministic when you need it.
Long-running when you don't.

Workflows are for processes that need to run the same way every time, or tasks that take hours or days. Agents figure things out. Workflows execute — predictably, repeatably, and under full control.

Visual Editor Drag & drop
Built by Analyst Agent
YESNO
Trigger
New contract uploaded
Action
Extract clauses
Branch
Risk score > 70?
Agent
Legal Reviewer
Notify
Slack #legal
Node palette
Trigger
Action
Branch
Agent
Notify
Built visually, or described in plain English — an agent creates the workflow for you.
Who builds them

No technical skills required

Drag & drop

Visual editor with nodes for triggers, conditions, actions, and agents. Connect blocks to build the flow — no code needed.

Agent-built

Describe what you want in plain English. An agent designs the workflow, you review and approve. For complex or repetitive processes.

API & YAML

For developers — define workflows as code. Version control, CI/CD, testable. Full parity with the visual editor.

Full control at every step

Stop, pause, retry — anytime

Workflows aren't fire-and-forget. Every step is controllable. Pause a long-running job to review intermediate results. Retry failed steps without restarting. Stop if something looks off.

Live execution — Contract batch analysis
Running · 2h 14m
Fetch contracts from Drive

2,403 files retrieved · 8.2s

Complete
Extract clauses

18,412 clauses identified · 47m 18s

Complete
Analyze risk scores Running
12,847 / 18,412
!
Call external API (rate-limited)

Failed: rate limit exceeded

Generate executive report

Waiting for previous steps

Pending
Zero-Trust by default

Workflows inherit permissions

Same rules as agents. A workflow invoked by a user runs with that user's scope. Triggered by an email? The sender's permissions apply. No accidental privilege escalation — ever.

Scope inherits from invoker

If Alice runs a workflow, all reads, writes, and actions execute as Alice. Even if the workflow touches sources Alice can't access, those steps are silently skipped — not escalated.

Email triggers map to user

Workflow triggered by an incoming email? The sender's identity becomes the execution scope. External senders get guest permissions automatically.

Philosophy

We enhance, we don't replace.

StellarBase workflows aren't trying to replace n8n, Make, or Zapier. Keep using the tools you love. Our workflows exist to power StellarBase — analyzing your knowledge, orchestrating agents, running long-form processing.

Call StellarBase from your existing automation. Trigger our workflows from anywhere. We integrate into your stack — you don't integrate into ours.

How it fits
n8n
Zapier
Make
Any webhook
triggers
StellarBase

runs the hard work

Example

Zapier detects new Gmail → triggers StellarBase workflow → analyzes attached contract → writes result back to Zapier → Zapier updates your CRM.

Chat

Chat where your team
already chats.

Bring StellarBase agents into Slack, Teams, Google Chat, or Discord. Or use our built-in chat — conversations, threads, notifications, multi-agent collaboration. Or skip the UI entirely and integrate via API into your own systems.

1. Bring agents to your existing tools

Install once. Mention anywhere.

Install StellarBase in your existing chat platform. Then just @mention any agent — they respond in-thread, inherit the channel's context, and respect user permissions.

1
Install in your platform
Slack Slack
Installed
Teams Teams
Installed
Google Chat Google Chat
Installed
D
Discord
Installed
+ any webhook-based platform
2
Just @mention in any channel
Hey @contract-reviewer
Agents 3 available
@contract-reviewer Legal docs analysis
@financial-analyst Q4 reports & metrics
@research-bot Web & internal research
, check this file |
Hover the mention above · type @ to summon any agent ↑↓ select insert
2. Or use our built-in chat

Full chat, built right in

Channels, threads, multi-user conversations, @mentions, notifications — the chat features you expect, with AI agents and live process tracking as first-class citizens.

legal-review · 14 members
StellarBase Chat
MK
Martin K. 2:14 PM

@contract-reviewer can you check this Acme SaaS agreement? Concerned about the liability terms.

Acme_SaaS_Agreement_v3.pdf · 24 pages
contract-reviewer StellarBase Agent 2:14 PM
Fetched document · 24 pages
Extracted 18 clauses
Comparing to your template library...
contract-reviewer 2:14 PM

Reviewed. 2 flags:

Liability cap set at €50K — your policy minimum is €200K. [§4.2, page 8]
Non-standard indemnification language — requires legal review. [§7, page 14]
All findings cited to source ·

Channels & Threads

Organized conversations

Notifications

Real-time pings, digest, mute

Multi-user + multi-agent

Humans and agents side by side

Live process tracking

See what agents do in real-time

3. Or build your own

Integrate directly via API.

For companies with their own internal tools, portals, or customer-facing apps. Embed StellarBase agents into any system through our REST API and WebSocket streams.

REST API for messaging & agent invocation
WebSocket streams for live responses
User-scoped tokens for per-employee auth
Webhooks for message events & notifications
curl
POST /v1/chat/invoke
Authorization: Bearer $TOKEN

{
  "agent": "contract-reviewer",
  "message": "Review §4.2",
  "context": {
    "doc_id": "a7f3...",
    "user": "alice@co.com"
  },
  "stream": true
}
Security & Privacy

Enterprise-grade by default

Four layers of protection. Zero-trust authorization at every level. GDPR compliant by design, with optional CMEK and comprehensive audit logs.

Encryption

AES-256-GCM at rest
TLS 1.3 in transit
Customer-managed keys (CMEK)
End-to-end encryption

Access Control

Zero-trust per-source authorization
RBAC: Viewer, Editor, Admin, Owner
SSO: SAML 2.0 / OIDC
Per-source permissions

Data Protection

Automatic PII anonymization
Audit logs (90 days – 1 year)
Data residency choice
GDPR subject rights support

Reliability

99.7% uptime SLA
Multi-AZ redundancy
Automated backups
Point-in-time recovery
Compliance
GDPR
SOC 2 Type II In progress
ISO 27001 In progress
CCPA
EU Data Residency
HIPAA-ready
Deployment

Your data stays with you.
Always. Everywhere.

Self-host for full control. Use StellarCloud for zero ops. Or mix and match. Whatever you choose, we never hold your data — it lives in your own storage, your own database, your own integrations.

Max control

Self-hosted

Run StellarBase entirely on your infrastructure. Air-gapped possible.

Docker / Kubernetes
Customer-managed keys
Air-gapped deployment
Your own GPU inference
Zero external calls possible
EU EU-only

StellarCloud

Fully managed by us. 2-minute setup. Data still lives on your side.

Zero ops, zero infra
AWS eu-central-1 (Frankfurt)
99.7% SLA
Auto-scaling
Bring your own storage & DB

Hybrid

Host yourself, but use StellarCloud, StellarGate, or your own model keys.

Self-hosted base
Cloud inference as needed
StellarGate for privacy proxy
Your own API keys (OpenAI, Anthropic...)
Mix and match anytime
Data sovereignty

Even in our cloud — your data never leaves you.

When you use StellarCloud, we process your data in transit, but we don't store it. Your connected storage and your database hold everything — we're just a processing layer.

Your side Storage, database, integrations
Our side Processing (no retention)
Back to your side Enriched data returns to you
Your Postgres
Your S3 / MinIO
Your Drive / Notion
encrypted
StellarCloud
Process & Enrich
✓ OCR, extraction, embedding✓ Agent inference✓ Search queries
No data retention
encrypted
Processed data to your Postgres
Knowledge graph to your storage
Agent outputs to your app
We process, you retain. Zero data stored on our side. Zero lock-in.
EU Full EU infrastructure

Our cloud runs entirely in the European Union.

All StellarCloud services — data processing, agent inference, storage buffers — are hosted in Frankfurt and Amsterdam. Governed by EU law. Audited for GDPR. Not subject to extraterritorial access requests.

EU data centers only (Frankfurt, Amsterdam)
EU legal jurisdiction, Czech HQ
GDPR compliant by design
No US Cloud Act exposure
Compliance & standards
GDPR
Compliant
EU Data Act
Aligned
DORA
Ready
NIS2
Ready
ISO 27001
In progress
SOC 2 Type II
In progress
🇨🇿
Czech company, EU operations HRSD s.r.o. · Based in EU, governed by EU
Bring your own

Every piece is swappable

Connect your own database, storage, model providers, and identity systems. We're infrastructure-agnostic — use what you already trust.

Your database

Connect your own Postgres — self-hosted, RDS, Supabase, or cloud

Your storage

S3, GCS, Azure Blob, MinIO, or any S3-compatible system

Your API keys

OpenAI, Anthropic, Google, Mistral — your account, your billing

Your identity

SAML, OIDC, Okta, Azure AD, Google Workspace, custom

Collaboration

Humans and agents,
working side by side.

StellarBase is built around teams from day one. Shared knowledge, multi-agent threads, zero-trust access, and full transparency into what everyone — human or AI — is doing.

Multi-user, multi-agent threads

Teammates and multiple agents in the same conversation. Mention any agent, delegate subtasks, review outputs — all in one thread with full history.

Zero-trust access, chained

Every agent, workflow, and search inherits the invoking user's permissions. Your team collaborates on shared knowledge without accidental data exposure.

Live process tracking

See exactly what agents are doing in real-time — current step, progress, expected completion. Pause, intervene, or redirect at any moment.

Shared context, shared work

One knowledge base, one source of truth. Everyone searches the same data, agents reuse team insights, discoveries propagate automatically.

Notifications & mentions

Get notified when an agent finishes, when someone mentions you, when a workflow needs approval. Inline in the platform or pushed to Slack, Teams, email.

Full audit trail

Who asked what, which agent responded, what sources were accessed, what actions were taken. Every interaction is logged and reviewable.

Get Started

Ready to unify your knowledge?

14-day free trial. No credit card required. Deploy in minutes.

View pricing