# Orionfold > Orionfold's flagship is Flow: Living Documents on your Mac. Ordinary files > keep their sources, recurring jobs and history together. Daytime AI edits > arrive for review; Night Shift runs within authority you set in advance. > A Morning Briefing shows the results, changes and receipts to keep or revert. > Base is free forever. Pro is $10/month or $96/year, with 10 Pro days included > and no credit card to use. Flow 1.7 brings Jobs into the Workbench beside your document. > Relay provides repeatable agent work with cost receipts. Arena evaluates local > AI. Orionfold is a member of NVIDIA Inception. ## Living Documents - [The Living Documents Manifesto](https://orionfold.com/manifesto/): Eighteen ideas for work that continues and judgment that stays yours, with interactive workflow illustrations, evidence and primary-source citations. - [Room for a Renaissance](https://orionfold.com/essay/): Manav Sehgal's essay on making more room for the researcher, writer and builder in each of us. The foundational essay, retained at its original canonical URL. - [Essays](https://orionfold.com/essays/): Manav Sehgal on working knowledge, human judgment and Living Documents. Two essays, each with its own canonical URL; both also appear in the Story RSS feed. - [The Work We Want to Keep](https://orionfold.com/essays/the-work-we-want-to-keep/): The flagship essay on personal AI, founder history and the judgment people retain. Follows a Customer Research brief through saved Jobs, source relationships and review. Detailed evidence is retained in the companion research notes. - [Research notes for The Work We Want to Keep](https://orionfold.com/essays/the-work-we-want-to-keep/notes/): Supporting studies, source links, Flow development observations and their limits. A companion to the flagship essay, not a separate essay. ## Flow (flagship, released for Mac) - [Flow](https://orionfold.com/flow/): Flow 1.7 Living Documents for Mac: inspect saved Jobs in Edit, Run and Relations; run useful work and review each changed part. Start with 24 useful documents. Interactive product illustrations explain the workflow. Pricing, FAQ, technical specifications and the direct download remain linked. - [Flow product tour](https://orionfold.com/flow/tour/): Explore Flow 1.7 through real product captures. Start with Night Shift, Living Documents and the six Settings screens, then explore writing, files, models and receipts. - [Flow: Night Shift](https://orionfold.com/flow/night-shift/): Schedule work while your Mac is plugged in and idle. Flow can be open or quit. Review a Morning Briefing and keep or revert changes. Base includes deterministic jobs; Pro adds local model-powered Overnight notes. - [Flow: Living Documents](https://orionfold.com/flow/living-documents/): Start with useful work offers 24 documents in four categories, with five featured choices. The website also presents seven curated document previews. File-bound charts and tables refresh from their sources; Jobs and collecting definitions stay with the document. Completed Jobs apply changes before review; Keep records a decision, Revert restores the selected output part, and Later leaves applied changes pending. - [Flow: Settings](https://orionfold.com/flow/settings/): Six screens: General, Documents, Models, Smart Routing, Evidence and Night Shift. See where models run, manage local models, and set a Night Shift window. - [Flow technical specifications](https://orionfold.com/flow/specifications/): System requirements, document formats, approval-led Agency, model routes, proof, privacy boundaries, plans (Base free forever, Pro at $10 a month or $96 a year), and qualified measurements for Orionfold Flow on Mac. - [Flow: Writing with AI](https://orionfold.com/flow/writing-with-ai/): Approval-gated AI editing (seven actions, one trust path), Expand with Sources (a model uses Flow's lookups to read the document and search your open folders, every lookup narrated live, bounded to 12 lookups, the write arriving as an approved diff with a receipt), an in-document tool strip that states scope, cost, and route before you click, and long documents improved part by part. - [Flow: Receipts and evidence](https://orionfold.com/flow/receipts/): Run receipts that read as one card (named checks with severity pills, an answer-first evidence score, the exact billed cost, the model binding, a timeline), an evidence gallery, and benchmarks that measure local models on your own Mac. - [Flow: Models and runtime](https://orionfold.com/flow/models-and-runtime/): Run models on your Mac or through configured cloud API providers. Flow names the route and model. Compatible local models are shared automatically; Manage models shows what is available. - [Flow: Documents and files](https://orionfold.com/flow/documents-and-files/): On-device search whose results are citations re-checked byte for byte, a table editor, 34 chart types and 20 diagram types drawn offline from plain text in the file (Visualize proposes one after a selected table, list, or paragraph, reviewed like any edit; a chart gallery editor picks a type by the question it answers), plain folders of plain Markdown that any app can read, and a resource popover that shows what the app is using. - [Flow for enterprise AI owners](https://orionfold.com/flow/enterprise/): All nine adoption patterns, each tagged Today (shipped in the app) or Direction. How document-native evidence, approvals, and receipts answer enterprise questions about allocation, data classification, attribution, guardrails, evidence, routing, curation, system of record, and capturing expert judgment with consent. - [Flow press kit](https://orionfold.com/flow/#press): Quotable facts for journalists and analysts: status, patent-pending statement, measured numbers with dates, and the press contact. ## Flow launch stories (first-person, 2026-08-22 to 2026-08-27) - [The Pit Crew That Never Touches the Wheel](https://orionfold.com/story/the-pit-crew-that-never-touches-the-wheel/): The launch story. One founder wanted an AI writing tool that cannot touch a document without permission; here is the race car built around that rule, and why every part of Flow starts from the driver's seat. - [The Edit I Never Approved](https://orionfold.com/story/the-day-i-stopped-trusting-invisible-edits/): A proofreader quietly changed code inside a note. That silent edit became Flow's first rule: AI may propose a change, but only you can save it. - [The Fan Was the Only Status Light](https://orionfold.com/story/where-your-ai-actually-runs/): A proofread ran a Mac hard for twelve minutes, but the app could not say which model was working or where the document had gone. Why Flow names the model, the route, and the cost on screen. - [The Faster Model Was the Bigger One](https://orionfold.com/story/the-fastest-model-on-your-mac/): Weeks on a small local model because it looked faster. One measurement on the same Mac proved the larger model was quicker. Why Flow benchmarks models where you run them. - [The Search Result That Moved](https://orionfold.com/story/a-search-result-is-a-citation/): The right search result landed on the wrong paragraph because the file had changed. Why Flow's search results are citations re-checked byte for byte. - [The Chart That Stayed Wrong](https://orionfold.com/story/charts-that-come-from-the-words/): A pasted chart stayed wrong for six weeks because it still looked finished. Why Flow draws charts from the words in the file, so they change when the numbers do. ## Buy - [Arena Field Edition](https://orionfold.com/arena/): The paid edition of the free Orionfold Arena. One installer turns an NVIDIA DGX Spark into a private AI lab with an AI research teammate (agents, domain models, and tests). It proves itself on first boot with a verifiable receipt and stays proven for a year, every update re-gated and re-receipted. The software stays free and open; what you buy is the assembled, proven team delivered turnkey. $349 founding for the first 25 licenses, then $499 one time with a 12-month kept-proven window; $149/yr renewal after year one. - [Relay](https://orionfold.com/relay/): A licensed product that runs your AI agents for real client work, all on your own machine. It shows what each client costs you, and none of your data leaves your computer. $349 founding for the first 25 licenses, then $499 one time; $149/yr renewal after year one. ## Flagship paths - [Flow](https://orionfold.com/flow/): Orionfold's lead flagship. Living Documents keep recurring work, sources and receipts with files you own. Review daytime edits; authorize a bounded Night Shift in advance and inspect its Morning Briefing. Native Mac, patent pending, released as a closed-source app. Base is free forever and includes deterministic Living Document jobs, Morning Briefing and Review. Pro adds model-powered Overnight notes and other AI actions. - [Relay](https://orionfold.com/relay/): The operations layer for repeatable agent work. Run agents and client workflows from one place, keep approval in the loop, and see cost per client. - [Arena](https://orionfold.com/arena/): Orionfold Arena is the cockpit for local AI: run, compare, score, and train models on your own desktop. Arena Field Edition delivers the whole stack on an NVIDIA DGX Spark as a turnkey, proven lab; Cortex memory and the fieldkit toolkit are the stack around it. - [Workflows](https://orionfold.com/workflows/): Private Agent Starter Kits. One local workflow with a trigger, runner, evaluation, approval loop, artifact output, and audit trail. - [Experts](https://orionfold.com/experts/): Offline domain models with benchmarks and playbooks for sensitive work. Patent Strategist is the live flagship pack. - [Advisor](https://orionfold.com/advisor/): A governed local AI advisor over your own documents: every answer names its exact source, unsupported questions get a refusal, and every check it passed is a receipt you can re-run. Live; the trained model is free on Hugging Face. - [Learn](https://orionfold.com/learn/): Books, field notes, the DGX Spark Field Lab, and build stories that teach the operating model. - [Become AI Native](https://orionfold.com/become-ai-native-business/): The starting point. Get the AI Native Business book free, then follow the path from the book to running your own private AI team. The simplest way in for a team that is trying to buy AI capabilities it could own instead. ## Relay resources - [Relay Guide](https://orionfold.com/relay/docs/): Step-by-step docs for using Relay: getting started, organizing customers and projects, planning work with chat agents, running workflows with human checkpoints, using packs and licenses, and watching runs and costs. - [Relay API](https://orionfold.com/relay/api/): The developer reference for the local HTTP API your Relay instance exposes. Every route is documented from the source that implements it, with real request bodies, response shapes, error cases, and worked examples. - [Relay Packs](https://orionfold.com/relay/memos/): The Relay Packs series. How Relay turns into a working business function through installable packs, with a pack for marketing, an agency bundle, a web designer pack, and industry verticals. - [Relay Demo](https://orionfold.com/relay/demo/): A hands-on demo of the Relay cockpit with sample data. Nothing in it touches a real workspace. ## Proof libraries - [Proof](https://orionfold.com/proof/): The receipts wall. Every capability claim links to a test you can run yourself and get the same result. It shows where the small local models already win, and where the big cloud models still lead. - [Receipts](https://orionfold.com/receipts/): Every receipt on its own page: a frozen test you can rerun, the run itself, and what we found. The proof, not the pitch. - [Trust](https://orionfold.com/trust/): How Orionfold products are built to be trusted: where your data flows (and where it does not), the license terms, the supply chain behind the software, and how you stay covered over time. Plain answers to the questions a careful buyer asks. - [Software](https://orionfold.com/software/): Flow leads the catalog, with Relay and Arena beside it. Proof, Advisor, Cortex, the API, apps, and developer tools provide the evidence and building blocks behind them. - [Models](https://orionfold.com/models/): Open-weight AI models tuned for patents, security, law, finance, medicine, and space math. Free to download and run offline. - [DGX Spark](https://orionfold.com/dgx-spark/): Built on the NVIDIA stack. The book, open models, and tools all proven on the NVIDIA DGX Spark, a small AI supercomputer that runs on one desk with no cloud. - [Books](https://orionfold.com/books/): Implementation playbooks. The AI Native Business book, AI research on the NVIDIA DGX Spark, and Field Notes. Built on working code. - [Story](https://orionfold.com/story/): Building in public. Honest notes on what we shipped, what broke, and what we learned. - [Letter](https://orionfold.com/letter/): A monthly letter from the founder. Plain notes on the bet behind Orionfold, with sources you can check. - [Adoption](https://orionfold.com/adoption/): What is shipped, what is being built, and what sponsors can fund next. ## More - [About](https://orionfold.com/about/): Manav Sehgal, the builder behind Orionfold. Twenty-five years building AI, from Xerox PARC to Alexa Conversations and Amazon Nova, now folded into one startup for the solo builder. - [GitHub](https://github.com/manavsehgal): Source code and open projects. - [Hugging Face](https://huggingface.co/Orionfold): Open-weight models to download. ## Optional - [Privacy](https://orionfold.com/privacy/): How we handle data. - [Terms](https://orionfold.com/terms/): Terms of use.