ar.io DocumentationAI Agents & LLMs
These docs are built to be consumed by AI coding agents, not just read by humans. Everything below is a plain-text file you can fetch, paste, or point an agent at — no scraping required.
Fastest path: point your agent at https://docs.ar.io/llms.txt.
It's a short index that links everything else, following the
llmstxt.org convention agents already look for.
Machine-readable entry points
| File | What it is | Reach for it when |
|---|---|---|
/llms.txt | A short index of the documentation with links and quick facts | Discovery — the entry point |
/SKILL.md | A procedural agent skill: SDK patterns, code recipes, URL conventions, constants | Generating working code |
/llms-full.txt | The entire docs corpus as one plain-text file (~1.2 MB) | Deep context, or RAG ingestion |
/sdks/<sdk>/llm.txt | Per-SDK reference text (see table below) | Working inside a single SDK |
llm.txt vs SKILL.md — what's the difference?
They solve different problems and are meant to be used together:
llms.txt/llms-full.txt/llm.txtare reference. They're documentation flattened to plain text so a model has the facts in context. Passive — they describe what exists.SKILL.mdis procedural. It follows the Agent Skills convention: task-oriented instructions, known-good code recipes, and constants that teach an agent how to actually build on ar.io — which SDK call to make, the correct ArNS URL shape, what to avoid.
A useful rule of thumb: load SKILL.md when the agent needs to write code, and
llms-full.txt when it needs to answer questions about the wider platform.
Per-SDK reference text
Each SDK ships its own flattened reference, which is much cheaper to load than the full corpus:
| SDK | Plain text |
|---|---|
| ar.io SDK | /sdks/ar-io-sdk/llm.txt |
| Turbo SDK | /sdks/turbo-sdk/llm.txt |
| Wayfinder | /sdks/wayfinder/llm.txt |
| ArDrive Core JS | /sdks/ardrive-core-js/llm.txt |
| CLIs (ArDrive CLI, ARIO Deploy) | /sdks/(clis)/llm.txt |
Give your agent somewhere safe to run
Agents learn fastest by doing, but you don't want an agent spending real money or writing permanent data while it figures things out. The Testnet Sandbox is built for exactly this:
- Runs the full ar.io stack — upload, payment, ArNS, and gateway — on Solana devnet.
- Funded by a faucet, so there's no real value at risk.
- Nothing is permanent: data is purged after ~3 days and never reaches mainnet Arweave.
One step needs a human. The ARIO faucet is GitHub-gated and its OAuth consent can't be completed headlessly. Have a person claim once to the wallet your agent will use — after that, uploading, funding, and buying names are all scriptable. See Agents and CI.
Skills and integrations
Several ar.io repositories ship agent tooling:
- ARIO Deploy — a Claude Code skill for deploying apps to the permaweb and updating ArNS. See what the skill does and add it to your project for setup.
ar-io-gateway-operator— a conformant Agent Skill for operating an ar.io node: health, indexing lag, ArNS resolution, data retrieval, and common pitfalls. Ships in the ar-io-node repo under.claude/skills/ar-io-gateway-operator/.ario-testnet-faucet— the faucet claim API and flow, for agents working against the Testnet Sandbox. Ships in thear-io-faucetrepo.
/SKILL.md above is itself a conformant Agent Skill — it carries name and description
frontmatter, so you can drop it straight into a project as .claude/skills/ario/SKILL.md.
Tools built into these docs
Every page carries agent-friendly affordances:
- Open in AI — send the current page straight to ChatGPT, Claude, Scira, or T3 Chat.
- Copy as Markdown — grab the raw source of any page for pasting into a prompt.
- Ask Arie — an in-docs AI assistant that answers from this documentation with citations.
Pointing an agent at ar.io
Load the index
Fetch https://docs.ar.io/llms.txt
so the agent knows what's available and where.
Load the skill for code generation
Fetch https://docs.ar.io/SKILL.md
for known-good patterns, constants, and URL conventions.
Narrow to an SDK
Pull the relevant llm.txt (table above) instead of the full corpus — smaller context, better
answers.
Run against the sandbox
Build and test on the Testnet Sandbox before touching mainnet.
Related
Testnet Sandbox
Free Solana devnet environment — the safe place for agents to run
ar.io SDK
TypeScript SDK for ArNS, gateways, and Solana program operations
Gateway API Reference
HTTP endpoints for data retrieval and ArNS resolution
Get Help
Ask in the ar.io Discord
How is this guide?