Open source · Self-hosted · Fully private

Your AI assistant,
on your hardware

Chat through WhatsApp, Telegram, Slack, or Discord. Automate your home, server, and daily life. Your data never leaves your machine.

ZilchOS agent task
you → Good morning! What's on my calendar today and are there any news stories I should know about?
 
📋 Morning briefing
─────────────────────────────────
1. Fetch calendar events
2. Search top news headlines
3. Summarise and reply
─────────────────────────────────
 
🔧 calendar_query → 3 events found
🔧 web_search → "top news today"
 
You have a 10am standup, 2pm design review,
and a 4pm call with Sarah. Top story today:
EU AI Act enforcement begins next month.
 
Delivered to Telegram · 0 data sent to cloud
6
chat platforms supported
100+
ready-to-use skill recipes
200+
AI models to choose from
0
bytes sent to ZilchOS servers
🔒

Completely private

Everything runs on your own computer, Raspberry Pi, or home server. Your conversations, files, and data never touch anyone else's servers — not even ours.

💬

Where you already are

Chat in WhatsApp, Telegram, Slack, Discord, Signal, or the built-in web UI. No new app to install. Just message it like you message a friend.

🤖

Actually does things

ZilchOS isn't just a chatbot. It can search the web, run commands, control your home, manage files, monitor cameras, and even write its own tools.

Watch ZilchOS do real work

A 20-second look at the web app: ask a question, ask it to do something and watch it act, approve a sensitive step, and answer when it needs a detail — all from one screen.

Everything you'd expect.
And a lot you wouldn't.

ZilchOS combines a capable AI brain with real tools for automation, monitoring, and memory — all running on hardware you control.

Autonomous agent with task planning

Just ask in plain chat — "send a WhatsApp to Priya that I'm running late" — and ZilchOS recognises it needs to act, hands the job to its agent, and does it (no special command). Give it a bigger goal and it breaks it into steps, builds a visible checklist, and works through them one by one — searching the web, reading files, calling APIs, running scripts. You watch progress in real time, approve any sensitive action before it runs, and answer if it needs a detail to continue — or press Esc to stop at any point.

Agent loop Live progress Just ask
📈

SaaS Growth Engine

A one-click preset turns ZilchOS into an always-on growth team for founders. It sources and scores leads against your ICP, runs outreach cadences in your voice (drafted for approval, stops on reply), repurposes content across channels, and schedules social posts — with a built-in pipeline view showing your funnel and your best leads. Pick it in setup and it installs the skills, suggests the integrations, and schedules the jobs for you.

Lead sourcing Outreach cadences Pipeline CRM
🎤

Voice input & output

Speak your message and hear the reply read back aloud. Works with any Whisper-compatible transcription server and any OpenAI-compatible TTS.

STT + TTS
🧠

Hybrid memory

Conversation history, a searchable knowledge graph, and a long-term memory file work together so ZilchOS remembers who you are and what you care about.

RAG + Knowledge graph
📸

CCTV & motion detection

Connect any IP camera via RTSP. ZilchOS monitors for motion, saves snapshots, and can alert you on Telegram or trigger a Home Assistant action automatically.

Camera integration
📱

Phone as a hardware node

Pair your Android phone via ADB. ZilchOS can read its camera, GPS, sensors, and screen — turning it into an always-on sensor for your assistant.

Android
🏠

Smart home control

Native Home Assistant integration. Control lights, switches, climate, and scenes just by asking. Trigger automations from agent tasks or camera events.

Home Assistant
🔧

Self-evolving tools

When ZilchOS encounters a task it has no tool for, it writes one — a Python, JavaScript, or Bash script saved to your workspace and reused in future runs.

Auto-synthesis
☁️

Cloud & infrastructure automation

Manage Kubernetes clusters, AWS accounts, and Terraform stacks just by asking. ZilchOS drives the command-line tools you already use — read-only checks run on their own, while anything that changes infrastructure waits for your approval. Step-by-step Linux and Windows admin handbooks are included.

Kubernetes AWS Terraform
🌐

MCP server support

Connect any Model Context Protocol server and its tools instantly become available to the agent — databases, APIs, third-party services, no code changes needed.

MCP
📅

Scheduled automation

Set tasks to run on a cron schedule — morning briefings, weekly summaries, email cleanup, server health checks. Set it once and forget it.

Cron
📣

Social publishing & scheduling

Post, read comments, reply, and like across Twitter/X, Facebook, Instagram, LinkedIn, and WordPress — or schedule one update across all your connected Buffer profiles at once: post now, queue it, or set a time.

Buffer scheduling
✍️

Full blogging

Write, list, edit, and publish posts on WordPress and Tumblr — as drafts or live — and create posts on Medium, all from a chat message.

WordPress · Tumblr · Medium
🔌

OpenAI-compatible API

Drop ZilchOS in as the backend for any app or tool that speaks the OpenAI API. Switch models from behind the same endpoint, locally.

API compatible
👥

Multi-user & multi-tenant

Each user gets their own isolated conversation history and spending budget. Role-based access keeps admin controls separate from regular chat.

RBAC
🔐

Encrypted storage

ZilchOS can store its RAG knowledge index and feedback data AES-256-GCM encrypted on disk — and because everything runs on hardware you control, there's no cloud copy to leak in the first place.

AES-256-GCM
Works everywhere you work

ZilchOS meets you in the apps you already use. No new app to download. No new habit to form.

💬 WhatsApp
✈️ Telegram
💼 Slack
🎮 Discord
📡 Signal
📟 IRC
🌐 Web UI
🔊 Amazon Alexa
🏠 Home Assistant
🐙 GitHub
📧 Email / Calendar
📷 IP Cameras (RTSP)
🐦 Twitter/X · LinkedIn
✍️ WordPress · Tumblr · Medium
📣 Buffer

Also exposes an ✦ OpenAI-compatible API — use ZilchOS as the backend for any tool or IDE that supports OpenAI, like VS Code, Continue.dev, or Cursor.

Your model, your choice

Point ZilchOS at any AI provider — or run entirely offline with Ollama. Switch anytime without rewriting anything.

Provider Vision Local / offline Best for
OpenRouterrecommended Easiest start — 200+ models, one API key
Ollamalocal Fully offline, Raspberry Pi, air-gapped
Anthropic Claude Opus / Sonnet — best reasoning
OpenAI GPT-4o, GPT-4o-mini, Azure
Google Gemini Gemini 2.5 Pro, Flash — fast & cheap
Groq, Together AI, LM Studio … Depends Any OpenAI-compatible endpoint

The built-in router automatically picks between a cheap model for simple replies and a capable model for complex tasks — saving cost without sacrificing quality.

Your data stays
on your machine

ZilchOS has no cloud backend, no analytics, no telemetry. When you ask it something, the only request that leaves your network is the one to your chosen AI provider — and even that can be eliminated with a local model.

  • 🔒
    No ZilchOS cloud There are no ZilchOS servers. Your messages never pass through us.
  • 💾
    Encrypted at rest Conversations, memory, and knowledge graph can all be AES-256-GCM encrypted on disk.
  • 🏠
    Runs on your hardware Raspberry Pi, old laptop, home server, VPS — if it runs Linux, it runs ZilchOS.
  • 🌐
    Works fully offline Pair with Ollama to run a local AI model. Zero external network calls required.

Data flow

You (Telegram)
Telegram API
encrypted
Telegram API
ZilchOS (your server)
your server
ZilchOS
AI Provider API
only the message
AI Provider
ZilchOS (your server)
answer only
ZilchOS
You (Telegram)
your server
With a local model (Ollama), the AI Provider step is removed entirely — your messages never leave your network.
100 things ZilchOS can do for you

A taste of what's possible — and ZilchOS ships with skill files for all of these, ready to activate.

🌅

Morning briefing

Calendar, weather, news headlines and a motivational note — delivered to Telegram before you're out of bed.

🛡️

Security monitoring

Watch your server logs, alert on suspicious login attempts, and summarise what happened overnight.

💸

Expense tracking

"I spent €45 at the supermarket" — ZilchOS logs it, categorises it, and gives you a monthly breakdown on request.

🧬

Health log

Track sleep, steps, meals, and symptoms. Ask for trends or share a report with your doctor.

🖥️

Server admin

Check disk space, restart services, tail logs, and get a health summary — without opening a terminal.

✍️

Content creation

Draft blog posts, social captions, email newsletters, and scripts — then push them to your CMS.

📚

Personal knowledge base

Save notes, articles, and ideas. Ask questions later and get answers pulled from your own library.

🕯️

Smart home scenes

"Movie mode" — dim the lights, close the blinds, set the thermostat — all from a single message.

Up and running in three steps

No compile steps, no Python environments, no complex config. One Docker command and you're chatting.

1

Copy the config file

Clone the repo and copy the example config. The setup page at /setup walks you through every setting with a visual model picker — no need to know model IDs by heart.

cp config.yaml.example config.yaml
2

Start with Docker

One command. No build step. ZilchOS is a single static binary with no runtime dependencies. It starts in under two seconds.

docker run -d --name zilchos \
  -p 8080:8080 \
  -v $(pwd):/data \
  zilchos:latest
3

Connect your apps

Open the web UI, add your Telegram bot token or other connectors, and start chatting. The admin panel handles everything — no YAML editing required for daily use.

http://localhost:8080
Ready to own your AI assistant?

ZilchOS is open source and free to self-host. No subscription, no usage limits, no data harvesting. Your hardware, your rules.

Runs on Linux · macOS · Windows · Raspberry Pi  ·  Docker or bare metal