Show HN: HART OS – an open-source AI OS built so frontier AI needs no datacenter HART OS, an open-source AI-native operating system from Hevolve, runs locally on any device without requiring a datacenter, providing an OpenAI-compatible API via a Model Bus that serves LLM, vision, and speech to all apps. The system, currently in public alpha, uses a single Python codebase that operates on laptops, servers, edge nodes, phones, and embodied AI, with peer-to-peer federation over WebSocket and boot-time security hashes. Hevolve Hive Agentic Runtime The AI-native operating system for every device, from your computer to embodied AI. Local-first, federated, OpenAI-compatible. HART= the bare engine pip install hart-backend , listens on :6777 .HART OS= the full AI-native OS. It boots on a laptop, server, or edge node, runs on phones, and reaches into embodied AI, and it ships the agentic Liquid Shell, Model Bus, model catalog, channel pairing, agent dashboard, and hive view.= the consumer companion app, one signed client across Windows / macOS / Linux. Nunba AI-native means the OS adapts to the machine, not the other way around. On each device it probes what the hardware can actually do, serves LLM, vision, and speech to every app over the Model Bus socket, D-Bus, or HTTP , and lets the on-device model compose the interface and learn each task once so it can replay it later. The runtime that drives a desktop is the same one that drives a robot, so a robot's AI access is just another Model Bus call. It is one Python codebase that runs in three shapes flat laptop, regional LAN, or central cloud mesh , speaks the OpenAI protocol on :6777/v1/chat/completions , and federates with peers over PeerLink direct peer-to-peer WebSocket, no broker . A boot-time guardrail hash, re-checked every 300 seconds, plus Ed25519 release signing, keep humans in control. You would notice it last, the way you notice anything alive: it improves on its own. Each node learns from what it does and gets quietly better, locally, on your own hardware, with nothing leaving the device. Calling an operating system alive should make you reach for the off switch, so that came first: the self-improvement is a toggle, every node is killable on its own, and it runs only as long as you let it. This README is written to be read by people and by agents alike. Every capability below names the file it lives in, so whether you are a developer or an AI agent exploring the repo, you can go from a feature straight to its source. Status: public alpha.The runtime, the Model Bus and the channel adapters are in daily use; APIs still move. Issues and PRs are genuinely wanted — see Contributing . If you would rather argue than patch, start atNine things we have not solved, each with the code that implements today's inadequate answer and what would count as progress: what convergence can mean with no global view, whether a system that rewrites itself can still be verified, and why a turn escalates itself to a better model automatically but can never decide on its own that a problem deserves an hour and three machines. Disagreeing with a framing there is worth more to us than a patch. Open problems . Why HART OS? why-hart-os Open problems /hertz-ai/HARTOS/blob/main/OPEN PROBLEMS.md — what we have not solved 60-second start 60-second-start How it compares how-it-compares Capabilities capabilities Hello, agent hello-agent Architecture map architecture-map API surface api-surface How auto-evolve works how-auto-evolve-works How hive connectivity works how-hive-connectivity-works Topology topology Build / extend build--extend Economics for node operators economics-for-node-operators Documentation index documentation-index License license Most software described as AI-powered ships an assistant: a separate app, usually talking to somebody else's server, that can drive a few functions. Remove the assistant and everything underneath works exactly as before. HART OS inverts that. Inference becomes a service the system provides, the way it provides a filesystem or a network stack. An application does not bundle a model or hold an API key — it asks the OS, and the OS decides which model answers, running locally where it can. Ten apps on one machine do not each load their own copy or each pay their own bill. That has a practical consequence worth stating plainly: every device becomes the same target. The runtime driving a laptop is the runtime driving a robot, so a robot's AI access is just another Model Bus call, and code written against :6777/v1/chat/completions runs unchanged on both. If you are here to contribute , the parts that most need outside eyes are the auto-evolve loop autoresearch loop.py , the guardrails that gate every self-improvement hive guardrails.py , and the 31 channel adapters — the most self-contained place to start. See CONTRIBUTING.md /hertz-ai/HARTOS/blob/main/CONTRIBUTING.md . git clone https://github.com/hertz-ai/HARTOS.git && cd HARTOS python3.10 -m venv venv && source venv/Scripts/activate Windows: venv\Scripts\activate.bat pip install -r requirements.txt echo "OPENAI API KEY=sk-..." .env or GROQ API KEY, or none for local llama.cpp python hart intelligence entry.py listens on :6777 OpenAI-compatible drop-in for any OpenAI SDK / LangChain / LiteLLM / Aider / Continue curl -X POST http://localhost:6777/v1/chat/completions \ -H "Content-Type: application/json" \ -d '{"model": "hevolve", "messages": {"role": "user", "content": "Hello"} }' Live demo https://hevolve.ai · Full quickstart https://docs.hevolve.ai/getting-started/quickstart/ · Nunba desktop https://github.com/hertz-ai/Nunba | HART OS | OpenAI Agents | LangChain | AutoGen | | |---|---|---|---|---| | Self-improves at runtime auto-evolve loop + RSI-2 gate | yes | no | no | partial | | Continuous baselining vs prior snapshots | yes agent baseline service.py | no | no | no | | Built-in benchmark adapters | 7 registry-driven | n/a | n/a | n/a | | Federates across peer nodes | yes PeerLink + hash-verified | no | no | no | | Local-first multimodal | yes llama.cpp + Whisper + 6 TTS + VLM | no | partial | partial | | Channel adapters out of the box | 31 | 1 webhook | custom | custom | | One codebase, multiple topologies | flat / regional / central | hosted only | library | library | | OpenAI-compatible endpoint | yes | yes | bring your own | bring your own | | Recipe replay cached LLM steps | yes 90% faster | no | no | no | | Native source protection HevolveArmor | yes | n/a | n/a | n/a | | What it does | Where | | |---|---|---| CREATE / REUSE recipe pattern | Run a task once via LLM, save the trace, replay 90% faster with no LLM calls on cached steps | create recipe.py , reuse recipe.py | GoalManager | Unified goal lifecycle, guardrail-gated state machine, escalation hooks | integrations/agent engine/goal manager.py | AgentDaemon | Autonomous tick loop, circuit breaker, frozen-thread detection | integrations/agent engine/agent daemon.py | SpeculativeDispatcher | Fast draft model answers first, expert agent takes over if confidence drops | speculative dispatcher.py | ParallelDispatch | ThreadPoolExecutor fan-out across SmartLedger tasks | parallel dispatch.py | SelfHealingDispatcher | Catches transient failures, retries with backoff + alternative providers | self healing dispatcher.py | 96 expert agents | Coding, research, marketing, product, security, ethics, ops, ... auto-dispatched per goal | integrations/expert agents/ | Recipe Pattern + Aider | In-process Aider backend no subprocess for code edits | integrations/coding agent/aider native backend.py | | What it does | Where | | |---|---|---| AutoEvolve loop | Realtime: hypothesis - 33-rule filter - hive vote - parallel sandbox - RSI-2 gate - federated broadcast | integrations/agent engine/auto evolve.py | RSI-2 monotonic gate | New release must beat prior baseline on every benchmark by configurable margin or PR is rejected | rsi trigger.py , pr review service.py | Benchmark registry | 7 built-in adapters 3 sourced from HevolveAI: QuantiPhy, Embodied, Qwen . Pluggable via register adapter | benchmark registry.py | Per-agent baselines | Per-agent snapshots at agent data/baselines/