Skip to main content
OpenPawz
Your AI, your rules. — The whole point.
The open-source AI agent OS built on Tauri v2. Create autonomous agents with 25,000+ integrations via MCP bridge, connect them to 11 chat platforms, and run everything locally on your machine. No cloud, no open ports, no Node.js backend.

Get Started

Install OpenPawz and bring up the app in minutes.

Create an Agent

Guided setup for personality, provider, and tools.

Connect a Provider

Add OpenAI, Ollama, Anthropic, or 7 other providers.

​
What is OpenPawz?

OpenPawz is the open-source AI agent operating system for your desktop. It’s not a chatbot. It’s not a wrapper around an API. It’s a full-blown, multi-agent platform that runs natively on your machine — with zero cloud dependency. Imagine this: You have autonomous AI agents with unique personalities, long-term memory, and access to 25,000+ integrations — managing Notion boards, Slack channels, GitHub repos, email, trading crypto on three chains, voicing responses aloud, monitoring infrastructure, controlling smart home devices, researching the web, deploying to AWS, writing content, and responding to messages across 11 chat platforms simultaneously. 400+ integrations ship built-in, and 25,000+ more are available via the embedded n8n engine and MCP Bridge — agents discover and install them at runtime, on demand. Just configure credentials and go.
Everyone who’s tired of walled-garden AI. Developers who want to build and extend. Power users who want a personal AI army across every platform they use. Teams who need private, auditable AI that never phones home. If you’ve ever wished your AI could actually do things — not just talk — OpenPawz is what you’ve been waiting for.

​
How it works

The desktop app is the single source of truth for agents, sessions, routing, and channel connections. Everything communicates through Tauri IPC — no open ports, no HTTP servers.

​
Key capabilities

25,000+ integrations

400+ native integrations built in, plus 25,000+ community integrations via the MCP Bridge to n8n. Agents discover and auto-deploy per-service workflows at runtime — zero configuration.

Multi-agent orchestration

Boss/worker delegation, persistent agent squads, inter-agent messaging, and workspace isolation. Full project lifecycle management.

Multi-channel gateway

Telegram, Discord, Slack, Matrix, IRC, WhatsApp, Twitch, and 4 more with a single app. Per-channel agent routing and smart message splitting.

Multi-chain DeFi

Coinbase + Solana (Jupiter/Pump.fun) + Ethereum (Uniswap). Trading policy engine with daily loss limits. Full trade audit trail.

Voice (TTS/STT)

Three TTS providers (Google, OpenAI, ElevenLabs) + speech-to-text. Auto-speak mode reads responses aloud as they arrive.

Visual node editor

Drag-and-drop workflow builder. Connect triggers, agents, tools, conditions, and outputs visually — no code required.

Model routing

Automatic complexity-based routing. Simple queries go to cheap models, complex tasks to powerful ones. Daily USD budget caps.

10 AI providers

Ollama (local & free), OpenAI, Anthropic, Google, DeepSeek, Grok, Mistral, OpenRouter, and any custom endpoint.

​
Memory that never forgets

Most AI tools forget everything between conversations. OpenPawz doesn’t.

Semantic long-term memory

Every important fact, preference, and detail is automatically captured and stored with vector embeddings. When context is relevant, it’s auto-recalled into the conversation — no manual prompting needed.

Memory Palace

A dedicated interface to browse, search, edit, and manage everything your agents have learned. Full visibility into what your AI knows about you.
Agents share a memory namespace by default, so knowledge captured by one agent is available to all. You can also isolate memory per agent or per workspace.

​
Token savings with session compaction

Long conversations burn tokens fast. OpenPawz has a built-in solution: automatic session compaction. When a conversation approaches the model’s context window limit, OpenPawz:
  1. Summarizes the older parts of the conversation into a compact digest
  2. Preserves key facts and decisions in memory
  3. Archives the full session for later reference
  4. Continues the conversation seamlessly with the summary + memory context
This means your agents can run indefinitely long conversations without hitting context limits or losing important context — and you pay for fewer tokens in the process.

​
Security you can actually trust

OpenPawz was designed from the ground up with a zero-trust, local-first security model.

100% local — no data leaves your machine

All conversations, memories, agent configs, and files stay on your hardware. No cloud sync, no telemetry, no analytics. The only outbound traffic is to AI providers you explicitly configure.

OS-level credential storage

API keys and secrets are stored in your operating system’s native keychain — macOS Keychain, Windows Credential Manager, or Linux Secret Service. Never in plain text. Never in a config file.

Command risk classification

Every tool and shell command is rated on a 1–5 risk scale. High-risk operations (file deletion, system commands, network access) require explicit user confirmation before execution.

Sandboxed execution

Run agent shell commands in isolated Docker containers with configurable resource limits, network policies, and filesystem restrictions. Agents can’t touch your host system.

Per-agent policies

Control exactly what each agent can do — which tools it can access, which channels it can use, what risk level it’s allowed to execute at. Full granular permission control.

Prompt injection defense

Built-in prompt injection detection analyzes incoming messages for manipulation attempts and flags suspicious content before it reaches your agents.

TLS certificate pinning

Custom rustls config pinned to Mozilla root CAs only — OS trust store excluded. Even a compromised CA on the user’s machine cannot intercept provider traffic.

Request signing & memory zeroing

Every outbound AI request SHA-256 signed for tamper detection. API keys auto-zeroed from RAM on provider drop via Zeroizing<String>.

​
Quick start

1

Clone and install

git clone https://github.com/OpenPawz/openpawz.git
cd paw
pnpm install
2

Build the Tauri app

pnpm tauri dev
The app launches with the built-in UI. No separate backend needed.
3

Add a provider

Open Settings → Providers and add an API key for OpenAI, Anthropic, or point to a local Ollama instance.
4

Create your first agent

Go to Agents → New Agent, pick a model, and start chatting. Your agent has immediate access to 25,000+ integrations via the MCP Bridge.
Need the full install guide? See Installation.

​
Start here

Installation

Prerequisites and build instructions for macOS, Linux, and Windows.

First Agent

Step-by-step guide to creating and configuring an agent.

First Provider

Connect to any of 10 supported AI providers.

Channels

Connect Telegram, Discord, Slack, and 8 more.

Integrations

Browse the 25,000+ integrations available to agents.

Architecture

Understand how the Tauri + Rust + TypeScript stack fits together.

​
Learn more

Agents

Agent configuration, policies, and lifecycle.

Memory

Semantic memory, auto-recall, and the Memory Palace.

Orchestrator

Multi-agent workflows and task delegation.

Browser

Headless browser automation and research.

Security

Command risk classification, sandboxing, and agent policies.

Troubleshooting

Common issues and diagnostic steps.