> Markdown twin of https://adna.network/ > Index: https://adna.network/llms.txt · Full corpus: https://adna.network/llms-full.txt > State is a build-time snapshot generated 2026-09-01 (UTC); nothing here is live. > Derived from the rendered page — this route has no markdown source. *[image: A pixel-art hero in the Tokyo Night palette: a glowing cyan DNA double-helix rising from a row of small computers and branching at the top into a network of connected nodes — shared inheritance resolving into the aDNA network.]* Open standard · MIT # The aDNA Network aDNA (agentic DNA) is an open standard for organizing a project's files, so that AI agents and the people working with them can always find what they need. Three folders, plain Markdown, tracked in git. This site is the standard itself, its docs, and the registry of the workspaces — 'vaults' — that run it. For teams working with agentic coding tools on real projects. Not a product or service — no server, no signup, nothing leaves your machine. Your context is just the notes, docs and decisions you already keep — now in a shape your agents can follow. The standard that gives it that shape is open. Your files stay on your machine. [Explore the network](/vaults) [Get Started](/get-started) $ `git clone https://github.com/aDNA-Network/aDNA.git ~/aDNA && cd ~/aDNA && claude` [Open source on GitHub](https://github.com/aDNA-Network/aDNA) · MIT-licensed · [the standard, versioned and public](/reference/specification) 74 vaults — every one of them on a single computer, ours. 15 are joined by 14 declared relationships; the rest stand alone. [The state of the network, dated →](/state-of-the-network/) [Who's behind aDNA →](/about/) 74 Vaults 16 [Entity Types](/learn/what-is-adna#entity-types) 3 [Conformance Levels](/glossary/glossary-conformance-level/) v2.5 Current Version MIT Licensed The connected aDNA network — 15 vaults joined by 14 relationshipsA compact radial map of the live aDNA network: 15 vaults are joined by 14 cited relationships (umbrella, federation, partner, companion) and drawn as a burst around the most-federated hubs (III and Astro), with the standard, aDNA, at the core. Each box names a vault and its persona; the full vault list at /vaults/graph is the keyboard-navigable twin. - aDNA.aDNARosettaIII.aDNAArgusWilhelmAI.aDNAHygieiaRareArchive.aDNAMnemosyneAstro.aDNAMolecules.aDNAFranklinTappProtocol.aDNAMentorVAAS.aDNAHarness.aDNAPanaceaOration.aDNARobert KennedyScienceStanley.aDNAZenZachary.aDNApygmalionVisualDNA.aDNApygmalionRemoteControl.aDNATalosTerminal.aDNAberthier 15 connected vaults · 14 relationships Language and DNA were co-created by everyone before us. The context that powers AI should be too — built, shared, and governed in the open, for the good of all. You already do the first half of this. The README that explains the project. The decision someone wrote down so nobody re-litigates it. The note on why the schema is shaped the way it is. That is context — aDNA gives it a shape your agents can read, in the open, in your own repository. ## What a context democracy is People and their agents keep their own project context, in the open, and no one company owns the whole of it. Each project keeps its own files. Projects cite each other, and what they choose to share becomes a commons anyone can read. **Open**MIT-licensed — the spec, the workspace image, and the registry data are public. - **Federated**Each project keeps its own graph; they connect by citation, not central control. - **Co-owned**No vendor owns your context — you govern it, in the open. - **Agents & humans**One graph, legible to your AI and to you at the same time. The aDNA network Real aDNA vaults — Astro, III, RareHarness, wga, RareArchive, Home — connected around the aDNA core by the relationships they declare. - Astro III RareHarness wga RareArchive Home aDNA the network The aDNA network Real aDNA vaults — Astro, III, RareHarness, wga, RareArchive, Home — connected around the aDNA core by the relationships they declare. Astro III RareHarness wga RareArchive Home aDNA the network Real aDNA vaults — forges, frameworks, platforms, and public-good archives — connected by the relationships they declare. ## The living registry Every vault is a real, governed context graph with its own place in the network. Most are tended by a named agent — an [AI persona](/about/#agent-stewards), not a person. Here's a slice across 74 of them. All of them run on one computer — ours. [The state of the network](/state-of-the-network/) says what that means. [the standard in use ### aDNA tended by Rosetta Open vault →](/vaults/adna/)[framework in use ### III tended by Argus Inspect/Introspect/Improve quality framework Open vault →](/vaults/iii/)[platform planned ### Canvas tended by Mondrian Open vault →](/vaults/canvas/)[coordination in use ### Operations tended by Berthier Open vault →](/vaults/operations/)[knowledge graph in use ### LA Venture Graph tended by Cartographer LA venture ecosystem knowledge graph (UCLA Anderson + Nerdstage/Demo Day LA) Open vault →](/vaults/laventuregraph/)[node in use ### Home tended by Hestia Open vault →](/vaults/home/)[org vault chartered ### wga World Genome Academy — buildpack + symphony + site Open vault →](/vaults/wga/)[org vault chartered ### RareArchive tended by Mnemosyne Rare Archive OSS rare-disease AI project Open vault →](/vaults/rarearchive/) [See all 74 vaults →](/vaults/) · [View the relationship graph →](/vaults/graph/) ## How it Works Three steps from scattered files to a project your agents can navigate — and keep navigating. **Agent-native** [Governance files](/glossary/glossary-governance-file/) and typed context give agents orientation — no prompt re-engineering every session. **Human-readable** Every file is plain Markdown. Browse in Obsidian, VS Code, or GitHub. No proprietary formats, no lock-in. **Composable** A *module* is one capability, a *dataset* is data it works on, and a *lattice* wires modules into a workflow. Start small, and grow to multi-team campaigns using the same three pieces. 01 ### Structure Every agent session starts from scratch — agents relearn your project by rummaging through files. aDNA ends that. Three directories (what you know, how you work, who's involved) give any agent instant orientation. aDNA.aDNA/ ``` aDNA.aDNA/ ├── CLAUDE.md ← agent operating protocol ├── STATE.md ← current phase, blockers ├── what/ ← what the project knows │ ├── context/ typed context library │ └── decisions/ architecture records ├── how/ ← how it operates │ ├── campaigns/ strategic initiatives │ └── missions/ decomposed work └── who/ ← who's involved └── governance/ roles & policy ``` 02 ### Orient Without a map, agents blast through irrelevant files or ask you to re-explain the project. Agents read your CLAUDE.md and [AGENTS.md](/glossary/glossary-agents-md/) first, then pull typed context at exactly the depth they need — not your entire repo, not a blank slate. aDNA.aDNA/CLAUDE.md ``` # CLAUDE.md — aDNA.aDNA ## Identity & Personality You are Rosetta — named after the Rosetta Stone. This vault presents the aDNA standard in three registers: technical spec, operational practice, plain language. ## Standing Orders 1. Phase gates are human gates. 2. Destructive actions require confirmation. ``` 03 ### Execute Context windows close and wipe progress — the next agent starts over. aDNA decomposes work into sessions, missions, and campaigns — context-sized chunks that fit a single agent window. What one agent learns, the next inherits. mission_wadna_p3_iterate.md ``` mission_id: mission_wadna_p3_iterate phase: 3 status: in_progress ## Decade backbone D1 Credibility-integrity — active D2 Navigation & docs — queued D3 Agentic + community — queued D4 Visual craft — queued ``` ### New to aDNA? Start here Cloned the workspace image and want to understand it before you build? The standard embeds in `.adna/`, and the guided path is here — a five-minute tour, the core concepts, and hands-on tutorials. [Learn aDNA](/learn) [What is aDNA? →](/learn/what-is-adna) [Tutorials →](/learn/tutorials) Dive deeper: [Convergence Model](/learn/concepts/convergence) · [Concepts & Tutorials](/learn) · [Reference Specification](/reference) ## The Standard aDNA is an open specification — MIT licensed, open to contribution, designed for extension. v2.5 MIT License Open Standard 16 Entity Types [Read the Specification](/reference/specification) [View on GitHub](https://github.com/aDNA-Network/aDNA) ## Built to be read by agents This site is itself an aDNA vault — the structure it documents is the structure that produced it. Each surface below is generated from the same source as the pages themselves, so it cannot drift out of step with them. [/llms.txt](/llms.txt) A curated index of this site, written to be read by an agent rather than rendered. [/llms-full.txt](/llms-full.txt) carries the whole corpus in one file. - Markdown twins Add `.md` to a documentation URL — or send `Accept: text/markdown` — and get the source instead of the page. 223 pages have one. - [/api/registry.v1.json](/api/registry.v1.json) The vault registry as versioned JSON, so the network can be queried as data instead of scraped out of HTML. ## Join the network The network is open — run a node, share a vault, and help [shape the standard](/reference/governance-model/). Mission-aligned subnetworks are taking shape around real public-good work. - [World Genome Academy](/commons/#featured) Genomics knowledge as a shared commons - [Context Commons](/commons/#featured) Agentic-context literacy, shared - [Wilhelm AI for the Undiagnosed](/commons/#featured) AI for people living with undiagnosed disease - [Rare Archive](/commons/#featured) Open-source rare-disease AI [See the commons](/commons/) [Join the community](/community/) aDNA for [Solo Developer](/use-cases/solo-developer/)· [Enterprise Team](/use-cases/enterprise-team/)· [Educator](/use-cases/educator/)· [Startup](/use-cases/startup/)· [Researcher](/use-cases/research-lab/)