TORBO BASE™

A local AI server that runs on your machine. Five providers through one gateway. 42 tools. Persistent memory. Six channels. Zero data collected.

Download for Mac Linux · Docker Source Code

macOS 14+ · Apple Silicon | Linux · Docker Open Source · Apache 2.0 Patent Pending · USPTO 2026

Capabilities

One Gateway. Everything AI.

Route any model through a unified local server with tools, memory, and channels built in.

5

AI Providers

Anthropic, OpenAI, Google, xAI, and Ollama local models — all through one endpoint.

42+

Built-in Tools

File ops, web search, code execution, shell commands, screenshots, and more. MCP host ready.

∞

Persistent Memory

Library of Alexandria — structured memory that persists across sessions, agents, and channels.

◈

Document Intel

Ingest PDFs, code files, and docs into a per-agent knowledge base with semantic search.

⚡

MCP Host

Native Model Context Protocol server. Connect any MCP-compatible client or build your own.

6

Channels

Telegram, Discord, Slack, Signal, WhatsApp, and the native macOS interface. One agent, everywhere.

Patent Pending — USPTO 2026

Six Tiers of Access Control.

The current answer to AI agent security is "use a separate computer." Ours is a six-tier access system that gates every tool, classifies every command, logs every request, and kills everything with one tap.

0

Off

All AI disabled. Kill switch.

1

Chat

Conversation only. No system access.

2

Read

Read files in sandboxed paths.

3

Write

Create and modify in allowed dirs.

4

Exec

Run approved commands and scripts.

5

Full

Unrestricted. Use with caution.

Interface

Built for Control.

Every panel designed for clarity. See what your AI can do, what it has done, and shut it down in one tap.

Dashboard — Access level dial, kill switch, system status

Dashboard Access level dial, kill switch, real-time system status

Agents — Permissions, access levels, capabilities

Agents · Permissions Per-agent access levels, tool allowlists, sandbox paths

Agents · Identity Personality, values, voice, and behavioral constraints

Skills — Code Reviewer, Document Writer, Web Researcher

Skills Reusable skill templates — code review, writing, research

Models — Installed and recommended Ollama models

Models Installed Ollama models + one-click install for recommended

Sessions — Conversation history across all clients

Sessions Conversation history across all channels and agents

Security — Access levels, sandbox paths, rate limiting

Security Six-tier access control, sandbox paths, rate limiting

Settings — API keys, voice, integrations

Settings API keys, voice settings, channel integrations

Get Started

60 Seconds to Running.

Terminal

# macOS — download and install
$ curl -fsSL https://torbo.ai/install.sh | sh

# Linux — run with Docker
$ docker run -d -p 4200:4200 ghcr.io/torbo-ai/torbo-base

# Add your first provider key
$ torbo config set anthropic.key sk-ant-...

✓ Torbo Base running on localhost:4200

Anthropic OpenAI Google xAI Ollama

Privacy

We Collect Nothing.

Zero Telemetry

No usage data leaves your machine.

Zero Analytics

No tracking scripts. No pixels.

Zero Crash Reports

Errors stay on your machine.

Zero Tracking

No cookies. No fingerprinting. No ads.

Your API keys are encrypted on disk. Your conversations stay on your machine. Your memory database is yours. We don't want your data — we want you to own it.

Something Bigger Is Coming.

Torbo Base is the foundation. What we're building on top of it will change how you work with AI forever.

No spam. One email when it's ready.