Build with Agents

Building on the Auki Network with an AI coding assistant? Give it the context it needs up front: a ready-made instruction set, plus a documentation map it can read on demand so it pulls the exact docs for the task — without you pasting anything into the chat.

Point your agent at the docs

The whole relevant documentation tree lives in a single llms.txt file (the llmstxt.org standard): every Auki Network repository's markdown — overviews, how-tos, references — listed as raw GitHub URLs, grouped by area, glossary first. Hand your agent the map and let it fetch what it needs.

🗺️ Full map: docs.dev.aukiverse.com/llms.txt — point your assistant here (e.g. "read https://docs.dev.aukiverse.com/llms.txt and follow the links you need").
📖 Glossary first: the Auki Network glossary — the vocabulary your agent should learn before anything else.

Grab the instructions for your tool

Auki maintains a single, cross-harness instruction set — following the AGENTS.md standard — that teaches your agent the Auki Network, the SDKs, the conventions, and how to run a local stack to build against. Download it and drop it into your project under the filename your agent reads.

Using the instructions with Claude Code

Click Download above — it saves as CLAUDE.md.
Claude Code loads CLAUDE.md from your project root automatically at the start of every session. For the full setup, also add .claude/settings.json and any skills under .claude/skills/.

Prefer the terminal?

curl -o CLAUDE.md https://raw.githubusercontent.com/aukilabs/auki-ai-wiki/master/AGENTS.md

One instruction set, every tool

Today this is one AGENTS.md that every popular harness can read (each just looks for it under its own filename), plus Claude Code's settings and skills — not yet separate per-tool skill bundles. The documentation map above is how your agent goes deeper.

Or clone the full AI workspace

The auki-ai-wiki repo is a clone-and-go workspace: the AGENTS.md instructions, reference repositories as submodules, and a local Auki Network cluster (DDS/DMS/console via Docker Compose) so your agent can build and test against a real stack without touching production.

git clone --recurse-submodules https://github.com/aukilabs/auki-ai-wiki.git
cd auki-ai-wiki
# then point your coding agent at AGENTS.md and bring up the local cluster
just up   # see the repo's justfile for the full workflow

Set up a domain to build against

Most of what your agent builds runs inside a domain — the digital representation of a physical space, anchored by printed portals. Before you test anything in the real world, set one up with the Domain Management Tool by following the domain setup guide on auki.com →: install the app, print or order portals, create the domain, pick a domain server and place the portals.

The Digital Twin Viewer Reference summarises the same steps and shows how the viewer then draws the domain.

Examples to learn from

Point your agent at a working project before asking it to build one. These are the reference builds to read, copy patterns from, and extend.

Main example: Domain Harness

aukilabs/domain-harness is a component-based spatial mapping prototype built on the Auki SDK. A Home server builds a shared scene from calibrated peer observations and serves a Three.js viewer; robot runners (Bracketbot) localize against portals and stream poses, cameras and voxel maps; an iPhone Annotator (Swift/ARKit + a Rust SDK bridge) aligns to a portal, places persistent annotations and contributes LiDAR depth. Peers talk over authenticated SDK Component protocols — a good end-to-end picture of networking, calibration and live spatial data in one repo.

https://x.com/broodsugar/status/2110905187896529081

Floorcraft: Blasters

matterless/floorcraft-blasters — the source of Floorcraft, a multiplayer shared-AR racing game in Unity: scan a QR code and you join your friends' arena in the same physical space.

Legacy stack

Floorcraft is built on the Unity ConjureKit SDK, which is deprecated. It's a good reference for shared-AR gameplay and QR-based session joining, but use the Auki SDK for new builds.

Helper repos and other examples

Repo	Use it for
auki-sdk	The SDK — helper functions, tools and docs for networking, calibration, marker detection and more
qr-lab	Just the QR code tooling
pnp-lab	Helpers for PnP calibration
experimental-modules	Authentication helpers — use the App Key / Secret from the console
domain-viewer	Example 3D web viewer for domain data
auki_robotics	Index of robotics integrations (general robots and Unitree G1)
auki_robotics_domain_calibrator	Calibrate a robot into a domain
auki_robotics_map_utils	Robot map utilities
auki_robotics_g1_humanoid_ros2	Unitree G1 humanoid ROS 2 integration
domain-spatial-computer	Spatial queries on domain data — pathfinding, raycast, navmesh projection
pathfinding	Navmesh/graph pathfinding library

What's inside

File	What it gives your agent
`llms.txt`	The documentation map — every relevant repo's docs as raw URLs
`AGENTS.md`	The core instruction set — Auki Network, SDKs, conventions
`.claude/settings.json`	Claude Code project settings
`docs/getting-started.md`	Human-readable getting-started walkthrough
`local-cluster/`	A full local Auki Network stack (Docker Compose) to develop against

tip

New to the Auki Network yourself? Read the Protocol section first, then let your agent take over with the map and instructions above.

Point your agent at the docs​

Grab the instructions for your tool​

Or clone the full AI workspace​

Set up a domain to build against​

Examples to learn from​

Main example: Domain Harness​

Floorcraft: Blasters​

Helper repos and other examples​

What's inside​