Open Assistant.
Your server. Your data.
A self-hosted personal assistant that connects your email, calendar, files, notes, and messaging into one conversational interface — no subscription, no vendor lock-in, no data leaving your infrastructure.
Your productivity tools are everywhere.
Your AI assistant shouldn't be.
Commercial assistants either live in the cloud with your data, only cover one ecosystem, or require you to paste context into a chat box. What you actually want is an assistant that can act — not just answer.
What's out there today
- ChatGPT doesn't touch your email or calendar
- Microsoft Copilot is locked to the M365 ecosystem
- Google Gemini only covers Google Workspace
- Self-hosted chat UIs have no action tools
- Workflow tools need manual pipeline building
- Everything costs $20–$30 / user / month
What Open Assistant gives you
- An assistant that already knows where everything is
- Cross-platform: Gmail and Outlook side-by-side
- Actions, not just answers — 88+ real tools
- Runs on your hardware, under your control
- Every action logged, every credential encrypted
- $0/month beyond your LLM API usage
Everything connected.
One conversation.
Email & Calendar
- Read, search and send via Gmail and Outlook
- Create and manage events in Google Calendar and Outlook Calendar
- Smart label and folder management
- Draft emails from natural language
Files & Notes
- Browse and search OneDrive and Nextcloud
- Create and update pages in Notion
- Download, read, and query file contents
- Unified cross-source search with hybrid ranking
Web & Search
- Privacy-first web search via Brave
- Full browser automation — navigate, click, type, scroll
- Vision-based page understanding via screenshots
- Structured content extraction for data analysis
Messaging
- Chat via web UI or WhatsApp
- Persistent conversation history with search
- Smart context management and auto-summarization
- Voice note processing and transcription
Automation & Scheduling
- Recurring tasks with cron expressions
- One-time future reminders
- Job execution history and monitoring
- Multi-step agent workflows, fully automated
Security & Transparency
- All credentials encrypted at rest (Fernet)
- Full audit log of every tool invocation
- OAuth2 for all service integrations
- Zero telemetry — no phone-home
One message. Nine specialized agents.
Your message is routed to one or more specialized agents — each with the right tools for the job. A Coordinator orchestrates multi-step workflows. You don't configure anything. You just talk.
You send a message
Find the invoice from Acme Corp last month, attach it to a reply to their latest email, and add a reminder to follow up in 3 days.
The coordinator routes it
Intent detection identifies the required agents: File Handler searches OneDrive and Nextcloud. Communication composes the reply. Planner schedules the follow-up.
Tools execute
Files are searched, the email is composed and sent, a future task is created — all in a single turn. Every action is written to the audit log.
You get a response
A plain-language summary of everything done, with confirmation of each step taken and a link to the audit log entry.
Works with the tools you already use.
Connect only what you need. Each integration is opt-in and configured from the Settings UI — no config files required.
+ more via the community plugin system — see below.
Built for people who care
about where their data goes.
Designed for self-hosters, privacy-conscious individuals, and EU organisations that need GDPR-compliant, fully auditable infrastructure.
Encrypted credentials at rest
All service credentials are encrypted with Fernet symmetric encryption before being stored. Raw tokens are never written to disk in plaintext.
OAuth2 everywhere
All integrations use OAuth2 flows. No raw passwords stored. Tokens are refreshed automatically and scoped to the minimum required permissions.
Full audit log
Every tool invocation is logged — what ran, when, with what arguments. Every settings change and authentication event is recorded. Nothing happens silently.
Zero telemetry
No analytics, no error reporting, no phone-home. The only outbound calls are to your chosen LLM provider and the services you explicitly configure.
Single-container surface
The entire application runs in one Docker container. The attack surface is contained, the codebase is auditable, and the deployment footprint is minimal.
Transparent plugin model
Community plugins are plain Python you read before running. No auto-updates, no opaque binaries. You decide what runs in your assistant.
All personal data stays on your infrastructure. Emails, calendar events, files, and conversation history are processed and stored exclusively on your own server. No data is transmitted to third parties beyond your chosen LLM API provider. Data residency, retention, and deletion are fully under your control.
Real workflows.
Fully automated.
From a single message, Open Assistant executes multi-step workflows across all your connected services — no manual glue required.
Extend it. Share it.
Own the code.
Build your own integrations in Python and share them with the community. Every plugin you install is readable source code — no black boxes, no auto-updates you didn't approve.
Write a plugin
Define a schema and a handler. The multi-agent router picks it up automatically and routes relevant requests to your new tool.
Share with the community
Publish to the open plugin registry. Other users can discover, inspect, and install your plugin — all from the Settings UI.
Registry, review guidelines and publishing docs coming with the first stable plugin API release.
Install with confidence
Every community plugin links directly to its source repository. You read the code before it runs. No bundled binaries. No hidden network calls.
Plugins run with explicit permission scopes and appear in the audit log like any built-in tool.
How it compares.
| Open Assistant | ChatGPT Plus | MS Copilot | Google Gemini | Open WebUI | AnythingLLM | |
|---|---|---|---|---|---|---|
| Self-hosted | ✓ | ✗ | ✗ | ✗ | ✓ | ✓ |
| Open source | ✓ | ✗ | ✗ | ✗ | ✓ | ✓ |
| Email — send & read | ✓ Gmail + Outlook | ✗ | M365 only | Gmail only | ✗ | ✗ |
| Calendar management | ✓ Google + Outlook | ✗ | M365 only | Google only | ✗ | ✗ |
| File management | ✓ OneDrive + Nextcloud | ✗ | OneDrive only | Drive only | ✗ | ✗ |
| Notion integration | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Web browsing | ✓ | ✓ | ✓ | ✓ | ✗ | ✗ |
| WhatsApp + voice notes | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ |
| Task scheduling | ✓ cron + one-time | ✗ | ✗ | ✗ | ✗ | ✗ |
| Full audit log | ✓ | ✗ | partial | ✗ | ✗ | ✗ |
| Community plugins | ⚡ coming | ✓ GPTs | ✗ | ✓ Extensions | ✓ | limited |
| Monthly cost | $0 | $20 / user | $30 / user | $20 / user | $0 | $0 |
LLM API cost not included — typically $5–20/mo depending on usage and provider.
Ready when you are.
We're putting the finishing touches on Open Assistant. Sign up to be notified when we launch, or reach out to discuss early access.
Designed for extensibility.
Clean layered architecture: API → Business Logic → Integration → Data. New tools are a schema and a handler — the router picks them up automatically.
- → 88+ tools across 9 specialized agents
- → Any OpenAI-compatible LLM — GPT-4o, Claude, Gemini, Ollama
- → SQLite by default — PostgreSQL optional
- → REST API for all operations
- → Agent definitions are database-driven and editable at runtime
- → Vanilla JavaScript frontend — no build step
- → Full docs: architecture, integrations, API reference
- → Plugin API coming — write your own tools in Python