Questions, answered.

The world is becoming agentic. Every week a new specialised AI agent appears — for research, writing, coding, trading, scheduling, shopping, therapy, fitness. Running them one by one doesn't scale, and none of them know what the others are doing.

SupraOS is the operating system for managing your agents. It gives you a full organisation of specialised agents that share memory, learn from outcomes, and get orchestrated by a central Co-CEO who delegates your intent to the right department. Departments include Co-CRO (research), Co-CTO (technology), Co-CMO (marketing), Co-CFO (finance), Co-COO (operations), and Co-CBO (business development).

What your agents can help you with, across every life area:

• Finance — portfolio rebalancing, DeFi yield hunting, P&L tracking, risk alerts
• Career — competitor intel, investor outreach, pitch decks, hiring pipelines
• Content — drafts, newsletters, social posts in your voice, engagement reporting
• Research — market scans, literature reviews, source extraction, sentiment trends
• Health — sleep/activity/HRV tracking, habit coaching, supplement planning
• Home & life — calendar triage, inbox sorting, travel booking, errand scheduling
• Learning — personalised curriculums, spaced repetition, progress reviews
• Relationships — follow-up reminders, gift planning, meaningful conversation nudges

Built by the Supra Labs team.

ChatGPT and Claude are powerful models, but every conversation starts from nothing. They forget you the moment the session ends, they live on someone else's servers, and you're locked into whatever model the vendor ships next.

SupraOS flips all of that:

• You own your data. Every memory, conversation, and decision lives in your own self-hosted store, encrypted with your keys. No mining, no training on your private context, no vendor data lake.
• It gets smarter the more you use it. Every interaction strengthens memory, links outcomes to decisions, and surfaces new institutional truths during the nightly dreaming cycle. Week one feels useful; month six feels indispensable.
• It runs locally. The daemon runs on your machine or a cloud instance you control. With local models (Llama, Mistral, Ollama), nothing leaves your hardware.
• No vendor lock-in. Swap models at any time — Claude, GPT-4, Gemini, Llama, Mistral, local. The new model inherits the full SupraOS state: memories, configs, reasoning chains. Nothing is stranded.

ChatGPT is a smart intern you re-hire every morning. SupraOS is a team that works for you, remembers everything, and stays yours.

Anyone who wants always-on AI executive assistance for every part of their life — a team of specialists working in the background 24/7, briefing you when it matters, acting when you delegate.

Founders running lean companies get the equivalent of a research analyst, a CFO, a CMO, and an operations lead from day one. Traders and DeFi operators get a finance desk scanning rates, portfolios, and alerts around the clock. Researchers and writers get a content team that knows their voice and sources. Busy humans get a personal assistant that actually remembers what matters — birthdays, investments, meal plans, health goals, home projects — and keeps the whole thing moving while they sleep.

Technically it's self-hosted, so you're most at home if you're comfortable running a local daemon — but the setup is a single install command and the interface is entirely conversational after that.

Four ways:

• Voice-first. Option+Space anywhere on your machine, or call/message an agent directly. Natural conversations, not prompts — "Hey Recon, what did our competitors publish this week?"
• Messaging apps. Reach agents through Telegram, Discord, Slack, or WhatsApp the same way you'd reach any collaborator. They remember the thread and respond with the same memory they use on the web.
• Scheduled workflows. Research every Monday at 6am, daily briefings at 7am, overnight synthesis. Tell the Planner agent what you need and it wires the cron + departments together.
• Triggered by signals. Competitor pricing changes, calendar events, hot inbox messages, portfolio thresholds, sensor data. SupraOS watches the world and activates the right agents automatically.

The daemon is entirely self-hosted. Your conversations, memories, and private keys never leave your infrastructure. Every agent action is cryptographically signed and hash-chained to a local audit log that only you can read.

Three tiers: department-level private memory (each team keeps its own), C-Level digests shared between department heads, and institutional truths confirmed by consensus across the organisation. Every memory is scored by MemoryRank — a 7-factor algorithm weighting access frequency, outcome correlation, recency, self-evaluation, cross-references, consensus, and human pin bonuses. The top-scoring memories fill the context window first.

No — memory decays with a 7-day recency half-life, then gets actively pruned during the nightly dreaming cycle. Useful memories (high access, good outcomes, cross-referenced) stay sharp. Dead-weight memories fade out. You can pin memories manually to make them immune to decay.

Every night SupraOS consolidates its memory. It merges near-duplicate memories, extracts patterns that repeat across departments, and promotes cross-departmental insights to institutional truths readable by every agent. The longer you run SupraOS, the sharper its memory becomes.

Yes — hindsight learning. When a past decision produces a bad outcome, SupraOS traces which memories were in play at the time and generates a "lesson" injected before similar future actions. Mistakes don't repeat, and wins compound.

Yes. ed25519 cryptographic identity for every agent, AES-256-GCM encrypted memory, SHA-256 hash-chained audit trail, self-hosted daemon. Agents can't mine your conversations — your keys are yours, stored locally. Turn SupraOS off and your data is gone; nothing retained on third-party servers.

The SupraOS daemon runs on your machine (or your own cloud instance). Your conversations, memories, and private keys never leave your infrastructure. LLM API calls go directly from your daemon to the provider of your choice — we never intermediate them. When you use Ollama, everything stays local.

Every agent action is cryptographically signed and chained to the previous entry (SHA-256 hash-linked, like a blockchain). Modify any single record and every subsequent hash becomes invalid — the chain breaks visibly. Periodic hash summaries are anchored to Supra L1 for external, decentralised proof no one can tamper with — not even us.

The Policy Engine. Every agent runs under a set of policies you set — what tools it can call, what addresses it can send to, how much it can spend per action / per day, which actions require a fresh approval from you (StarKey signature, email confirm, or in-app tap), and which are auto-approved.

• Per-agent delegation — "Co-CFO may transfer up to $500/day between my own vaults; trades over $50 need my approval."
• Spend caps — daily and per-action limits on both on-chain transfers and LLM API costs.
• Action whitelists — agents can only invoke tools explicitly granted to their role.
• Approval flows — high-stakes actions pause and ping you for sign-off before executing.
• Liquidation safeguards — thresholds that automatically unwind positions if something goes sideways.

Policies are enforced by IronClaw, the 10-module security framework sitting under every agent action. Nothing executes without passing its checks.

Yes. SupraOS is model-agnostic. Llama 3 and Mistral are included free. Claude, GPT-4, Gemini, and local Ollama models are plug-and-play. When you switch providers, the new model gets a full SupraOS state summary — memories, configs, reasoning chains, active context. Nothing is lost.

REST API, SDK (TypeScript + Python), MCP server, Webhooks, Telegram, Discord, Slack, WhatsApp, Gmail, Calendar, GitHub, Obsidian. On the financial side: wallet connections (StarKey, MetaMask, Phantom), exchange keys (Binance, Coinbase, Kraken and friends), DeFi protocol scanners. More being added weekly during the alpha.

Yes. @supraos/sdk ships with every install. You can extend any agent with custom tools, define new workflows, subscribe to memory events, and build your own agents that plug into the department hierarchy.

SupraOS anchors periodic hash summaries to Supra L1 — external, decentralised proof that your audit chain hasn't been tampered with. You can verify any agent action against the chain after the fact. 500K TPS, 0.4s finality, native oracle integration.

Mostly yes — if you're running Ollama for inference. The agents, memory, workflows, and audit chain all run locally. You lose the external LLM providers and the Supra L1 anchor round-trips, but the core system keeps working.

Anywhere Node.js runs. Mac, Linux, Windows. Self-hosted on your laptop, a home server, a VPS, or a cloud instance you control. Docker and Kubernetes configurations ship with the alpha.

Join the alpha launch — we're accepting 100 applications. Once approved, install runs via npx @supraos/create my-org and spins up your organisation in about 30 seconds.

Alpha is free for the first 100. Post-alpha pricing will be announced during the alpha period. OSS models (Llama, Mistral, Ollama) run at zero API cost; you only pay for premium models (Claude, GPT-4, Gemini) at pass-through pricing with no SupraOS markup.

Alpha means it's functional but evolving. Early users should expect regular updates and some rough edges. We're shipping rapidly — daily builds are common, and alpha users get direct access to the team for feedback.