v1.4.0 — Now Available for macOS & Windows | TigrimOSR (Rust Edition) Now Open Source

AI Agent
Powerhouse

Self-hosted AI desktop application with parallel multi-agent orchestration, cross-machine remote agents, secure sandboxing, and a skill marketplace. Agents self-improve during every run — human feedback auto-creates new skills so your agents get smarter with each interaction. Deploy agent swarms across your network — your data never leaves your infrastructure. No Docker required.

Download for Free View Source Code
TigrimOS v1.4.0
TigrimOS Task Manager with running agent tasks and Minecraft visualization TigrimOS AI Chat with tool-calling and React/Recharts visualizations TigrimOS Visual Agent Editor with drag-and-drop multi-agent design TigrimOS Live Agent Diagram showing orchestrator with active sub-agents and real-time status TigrimOS Human Feedback auto-creating skills — agent improves itself during run
Live Agent Diagram — real-time visualization of orchestrator and sub-agents with status, tool calls, and communication edges.
16 Built-in Tools
7 Agent Topologies
0 Cloud Dependencies
MIT Licensed
Enterprise Developers Indie & Solo Devs Engineering Teams Open Source Enthusiasts AI Researchers Regulated Industries
Sound Familiar?

The Problems You Face Every Day

Cloud AI tools are powerful — until they hit the wall of privacy, cost, and control.

Compliance blocks cloud AI

Regulated industries can't send sensitive data to third-party servers.

TigrimOS runs 100% on-premise inside an isolated sandbox. Zero data leaves your machine.

API bills that spike without warning

Multiple AI subscriptions stack up and blow through budgets overnight.

Run free on local models, or mix cheap local workers with cloud orchestrators. You control every dollar.

Locked into a single AI vendor

Switching providers means retraining workflows and losing integrations.

Assign different providers per agent — OpenAI, Claude, Ollama, Codex — swap anytime, no lock-in.

No offline or air-gapped option

Cloud AI doesn't work without internet — useless in secure or remote environments.

Full functionality offline with local LLMs. Works air-gapped — no internet required.

Can't customize or extend your AI tools

Closed-source tools can't be adapted to your domain or workflow.

MIT licensed. Build skills, connect MCP servers, export YAML workflows — own everything.

Hard to prove ROI to leadership

Per-seat AI licenses are expensive and first to get cut when budgets tighten.

Free and open source. No per-seat fees. Deploy to the entire team at zero licensing cost.
Save Your Budget

Flexible AI — Your Models, Your Cost

Choose the AI provider that fits your budget. Use premium cloud APIs when you need top performance, or run completely free on local models.

$0 / month

Run Local LLMs

Use open-source models on your own hardware. Zero API costs, zero data sharing, full privacy.

  • Ollama — run Llama, Mistral, Gemma locally
  • llama.cpp — lightweight C++ inference
  • LM Studio — GUI model manager
  • Air-gapped — works offline
Most Flexible
Mix & match

Hybrid — Best of Both

Assign different providers per agent. Use a powerful cloud model for planning and a free local model for execution.

  • Orchestrator on GPT-4 / Claude Opus
  • Worker agents on local Llama / Mistral
  • Claude Code CLI + Codex CLI as coders
  • Pay only for what you use
API pay-as-you-go

Cloud APIs

Connect any OpenAI-compatible API for maximum capability. Use your own keys — no markup, no middleman.

  • OpenAI (GPT-4o, o3, o4-mini)
  • Anthropic (Claude Sonnet, Opus)
  • Any OpenAI-compatible endpoint
  • Your keys — direct billing, no markup
Capabilities

Everything You Need to Run AI In-House

A complete workspace where your agent teams build, reason, and execute — without ever leaving your infrastructure.

Parallel Multi-Agent Orchestration

7 orchestration topologies and 4 communication protocols. Agents work simultaneously on different parts of your problem — mesh networking, P2P swarm governance, and YAML-exportable workflows.

16 Built-in Tools

Web search, Python execution, React rendering, shell commands, file operations, skills, sub-agents — all sandboxed and ready to use out of the box.

Mix Any AI Provider

Assign different models per agent — OpenAI-compatible APIs, Claude Code CLI, Codex CLI, or local LLMs via Ollama, llama.cpp, and LM Studio. One team, many brains.

MCP Integration

Connect any Model Context Protocol server — Stdio, SSE, or StreamableHTTP. Extend your agents' toolbox with external services, databases, and APIs.

Built-in Terminal

Full xterm.js terminal with root access to the Ubuntu sandbox. Install packages, manage services, run CLI tools — all from the browser with color, tab completion, and cursor support.

Skills & ClawHub Marketplace

Install pre-built AI skills from the marketplace or create your own. Package domain-specific capabilities and share them across your organization.

Human Feedback Auto-Create Skill

Rate agent responses with thumbs up/down and the agent automatically creates new skills from your feedback. The agent improves itself during the run — no retraining needed.

Long-Running Stability

Smart context compression, sliding window management, and checkpoint recovery. Your agents keep working through complex, multi-hour tasks without losing context.

How We Compare

TigrimOS vs Other Agentic Frameworks

See how TigrimOS stacks up against popular agentic frameworks across security, deployment, and orchestration capabilities.

TigrimOS vs Agentic Frameworks Radar Comparison
Agent Architecture

7 Ways to Orchestrate Your Agents

Choose the right topology for every task. Click each pattern to see how agents connect and collaborate.

A B C D Orchestrator Coder Researcher Reviewer EVERY AGENT CAN REQUEST ANY OTHER FOR HELP Plan Code Test Ship Design Implement Review Deploy ONE ROUTE: DESIGN → CODE → CHECK → SHIP HUB W1 W2 W3 W4 W5 CENTRAL HUB DELEGATES TO ALL WORKERS TASK A bid $2 B bid $5 C bid $1 D bid $3 WIN C wins task E AGENTS BID ON TASKS — BEST FIT WINS THE JOB SEND to all R1 R2 R3 R4 ONE AGENT SENDS THE SAME TASK TO ALL CEO LEAD WORKER M L1 L2 W1 W2 W3 W4 TREE LAYERS — MANAGERS DELEGATE DOWN STAR PIPELINE MESH ORCH A B C P1 P2 P3 COMBINES STAR + MESH + PIPELINE IN ONE SWARM

Mesh Network

Every agent can talk to every other agent directly. Any node can request help from any peer — no bottleneck, full redundancy. Best for collaborative problem-solving where context is shared.

Web Service Deploy

Deploy Agentic AI to Your Organization

Set up TigrimOS as a self-hosted web application. Give every team autonomous AI agents — from business ops to R&D — all on your own infrastructure.

Software Development

Agent teams that code, review, test, and deploy in parallel. Claude Code + Codex CLI as autonomous coders inside your swarm.

Research & Data Science

Run experiments, analyze data, and synthesize literature with full Python/numpy/pandas. Sensitive data never leaves your servers.

Business & Management

Automate reports, process documents, and run strategic analysis. AI agents handle routine work so your team focuses on decisions.

Tech Company & Enterprise

Full data sovereignty, air-gapped deployments, and org-wide skill sharing. MIT licensed — no vendor lock-in, no per-seat fees.

1
Download & launch — sandbox provisions automatically
2
Connect AI models — cloud APIs, local LLMs, or both
3
Design agent teams — visual editor, YAML export, ship faster
Distributed Intelligence

Cross Machine Remote Agents

Extend your agent swarm beyond a single machine. TigrimOS instances connect and collaborate across your network — LAN, VPN, or WAN — via a simple REST API with Bearer token authentication.

MACHINE A 192.168.1.10 Coder Tester LOCAL LLM (Ollama) MACHINE B 192.168.1.20 Data Analyst Writer MACHINE C 10.0.0.5 (VPN) GPU Worker MACHINE D cloud-gpu.example.com Fine- Tuner Deploy Agent CLOUD API (OpenAI) HTTP REST + BEARER TOKEN AGENTS SPAN MACHINES — EACH NODE RUNS ITS OWN SANDBOX & LLM

Connect Any Instance

Point any TigrimOS instance at another over the network. Submit tasks via REST API with a Bearer token — works across LAN, VPN, or public internet.

Token-Based Security

Each instance generates a unique remote token. All cross-machine calls are authenticated — no complex certificates, just enable Remote Mode and share the token.

Heterogeneous Hardware

Mix Mac, Linux, and Windows nodes freely. GPU-heavy inference on a rack server, coding agents on a laptop — orchestrated as one swarm.

Real-Time Progress

Poll running tasks for live status updates, progress logs, and elapsed time. Build dashboards or automation scripts on top of the simple polling API.

Full Agent Power Remotely

Remote tasks run through the same multi-agent orchestration as local tasks. All 7 topologies, 16 tools, and model configs are available across machines.

Sandbox Isolation

Remote tasks execute inside the same sandboxed environment as local tasks. The remote caller never gets host access — only task results come back.

1
Enable Remote Mode in TigrimOS Settings and copy the generated token
2
Submit tasks from any client — POST /api/remote/task with your Bearer token
3
Poll for results — get live progress, status, and final output
Zero Trust

Secure by Design

Every AI operation runs inside an isolated Ubuntu sandbox. Your host system is protected by default.

Full Sandbox Isolation

macOS: Virtualization.framework VM. Windows: WSL2. AI code physically cannot access your host.

Invisible by Default

Host files hidden from AI. Share only what you choose — read-only by default.

NAT Networking

Sandbox isolated from host network. No unexpected outbound connections.

No Docker Required

Native OS-level virtualization. Lighter, faster, zero Docker dependency.

Get Started

Download TigrimOS

Free and open source. MIT licensed. No account required.

macOS

Apple Silicon & Intel

macOS 13.0+ (Ventura) 4 GB RAM, ~5 GB disk Virtualization.framework
Apple Silicon (M1-M4) Intel

Windows

Windows 10/11

Windows 10 v2004+ 4 GB RAM, ~5 GB disk WSL2 sandbox
Download for Windows

Linux

Web-Based Agent Platform

Ubuntu / Debian / any distro Node.js 18+, 4 GB RAM Run as web service
View on GitHub

Build from Source

MIT Licensed

Full source code Swift + Node.js Community contributions welcome
View on GitHub

TigrimOSR

Rust Edition

Rewritten in Rust Lower memory footprint Cross-platform native binary
View on GitHub
Build from source 
# Clone the repo
git clone https://github.com/Sompote/Tigrimos.git
cd TigrimOS

# Install qemu (macOS)
brew install qemu

# Build
swift build -c release
./Scripts/build.sh silicon  # or: intel, all
Our Service

AI Agent Transformation for Your Organization

We help companies adopt AI agents to save time and reduce costs. From strategy to deployment, we bring autonomous AI into your workflow — tailored to your industry.

Engineering & Design

Automate design reviews, documentation, and technical analysis.

Supply Chain & Procurement

Streamline supplier evaluation, purchase orders, and logistics tracking.

Research & Development

Accelerate literature reviews, data analysis, and experiment tracking.

Business Operations

Automate reports, process documents, and run strategic analysis.

Software Development

Agentic coding — AI agents that code, review, test, and deploy in parallel.

Government & Highway

Agentic road survey, infrastructure inspection, and compliance reporting.

Professional AI Chatbot

Organization-tailored chatbot for internal knowledge and daily workflows.

Academic & Education

Agentic tutoring, curriculum QA, and auto-generated lecture materials.

Healthcare & Medical

Clinical documentation, patient record summarization, and literature review.

Legal & Compliance

Agentic contract review, regulatory monitoring, and risk flagging.

Manufacturing & QC

Production monitoring, defect analysis, and predictive maintenance.

Human Resources

Agentic recruitment screening, onboarding, and HR policy chatbot.

Ready to bring AI agents into your organization?

Contact Us

Stop Sending Your Data to the Cloud

Deploy autonomous AI agent teams that run entirely on your infrastructure. Free, open source, and secure by design.

Get TigrimOS Free Star on GitHub