A real MachinaOS instance is running on our server at demo.machinaos.ai. Start a private 30-minute session and see a repo workflow run end-to-end — no account, no API key, no installation needed.
Public demo runs on OpenAI gpt-4o-mini for speed. The desktop product is bring-your-own-LLM and supports 7 providers — including fully local Ollama so nothing ever leaves your machine.
Open the demo in Code Review mode and you should reach the core product loop in under a minute: a seeded workspace, a generated plan, real tool usage, and a structured result you can inspect.
You should see a live plan, active tools, and a concrete output quickly. If you want a narrated route through the UI, use the storyboard instead of exploring blind.
Use these entry links to land directly in a pre-loaded workflow instead of the generic demo home screen.
Lands in the Studio with a Code Review chain pre-loaded against the seed workspace. MachinaOS will analyse the files, flag issues, and produce a structured report.
Start Code Review →Triggers the Security Sweep composite chain — scans files for common vulnerabilities, cross-references the tool registry, and outputs a Production Readiness report.
Start Security Audit →Runs the Daily Brief chain — summarises recent Git activity, open issues, and outstanding tasks in the seed workspace and delivers a concise morning report.
Start Daily Brief →Know exactly what to type, what to expect, and why each moment is worth pausing on. Covers project exploration, the Tools view, multi-step agent plans, the Visual Studio canvas, and multi-agent coordination.
Includes the full suggested-prompt table and a presenter checklist for live walk-throughs.
Open the Storyboard →The storyboard is the shortest path to understanding the product surface without guessing which tab to open or what to type first.
status, log, diff against the seed workspace.fetch, memory, sequential-thinking, and demo are pre-connected.shell.run and process.stop are disabled for safety.Technical evaluators and integration partners can request a 24-hour reviewer session. Reviewer mode unlocks additional capabilities — including browser introspection and extended note persistence — while keeping the shared-infrastructure safety boundaries in place.
Contact us with your name and use-case and we will send a redeem link.
Journalists and investors covering MachinaOS can request a 7-day extended session for in-depth evaluation. The same safety boundaries apply; you get a persistent workspace across multiple visits.
Contact us and reference "press" or "investor" in your message.
The live demo runs on our shared server with sandboxed restrictions. For full capabilities — your own LLM provider, your real workspace, unrestricted shell — install MachinaOS locally.
Windows 10/11 (x64), Ubuntu 20.04+ (x64 / arm64), macOS 12+ (Intel and Apple Silicon). The desktop installer bundles the Python runtime — no separate installation required.
For fully local operation, install Ollama and pull a model (e.g. ollama pull llama3). For cloud providers, obtain an API key from OpenAI, Anthropic, Google, or OpenRouter before starting.
Approximately 200 MB for the MachinaOS desktop app. Local LLM models require additional space (Llama 3 8B ≈ 4.7 GB). The runtime stores logs, memory, and secrets in your user profile directory.
Download the self-contained installer for your platform. No Python, pip, or other prerequisites required — the runtime is fully bundled.
# Windows — run the MSI or NSIS installer # Linux — install the DEB or AppImage # macOS — open the DMG and drag to Applications
Or start the backend server directly for headless or API-only use:
curl -sL get.machina.os | bash
Open Settings and choose from seven providers. Ollama runs fully local on your machine; OpenAI, Gemini, Claude, OpenRouter, and LM Studio require an API key.
# Example: point to a local Ollama instance Base URL: http://localhost:11434 Model: llama3
Register your repository so MachinaOS can resolve file paths, tool availability, and process context before planning.
machina workspace connect --path ./project-x
Apply your policy profile for approval gates, execution limits, and logging requirements.
machina profile apply --name local-safe-default
When setup is done, run your first goal and review the plan preview, approvals, tool outputs, and timeline — all on your own machine with full capabilities.
machina run "Prepare my development environment for Project X"
Check that no firewall is blocking the local socket port (default: 8765). On Linux, verify AppImage execute permissions (chmod +x MachinaOS.AppImage). On macOS, allow the app in Security & Privacy if macOS blocks unsigned apps.
For Ollama: confirm the server is running (ollama serve) and the base URL matches (http://localhost:11434). For cloud providers: verify the API key is correct and the selected model name matches the provider's available models. Check the diagnostic error message in Settings — it distinguishes connection refusal, timeout, and auth failure.
Your active policy profile may be set to "auto-approve". Open Settings → Policy and select a profile that includes approval gates, or create a custom profile with require_approval: true for tool categories you want to gate.
Ensure the path you registered is absolute and the directory exists. On Windows, use forward slashes or escaped backslashes. Re-register with machina workspace connect --path "C:/your/project" and check the workspace list in the Explorer view.
Open the Studio view (Ctrl+6), drag tools onto the canvas, connect output ports to input ports, and click Save then Execute. Try the Debug button to step through one node at a time.
Open the MCP view, expand Quick Pick, and register a curated server like memory or fetch in two clicks. Discovered tools become first-class entries in chat, Studio, and chains.
Run a multi-step task and open Agents \u2192 Communications. The SVG graph shows requests, negotiations, handoffs, and broadcasts as color-coded edges between agent nodes.
Read the full features overview to understand planning, security, agent orchestration, the Studio Debugger, chain composition, and MCP integration in depth.
Browse user stories to see how developers use MachinaOS for repo onboarding, debugging, security audits, and multi-agent coordination in practice.
Check the interface gallery to preview all 14 runtime panels \u2014 from the Neural Link and Workflow Studio to the Security Center, MCP view, and Event Inspector.
No. The public demo creates an anonymous session automatically, so you can evaluate the core workflow before setting up anything locally.
Start with the Code Review quick start. It is the fastest path to a visible plan, active tools, and a concrete result inside the seeded workspace.
The local runtime lets you connect your own codebase, choose any supported LLM provider, and use capabilities that stay restricted in the shared demo environment.