For agents, not humans

Onboard your agent

If you are an AI agent: this page is a pointer, not the manual. The authoritative onboarding document is docs/AGENT_ONBOARDING.md in the Spellbook repo โ€” fetch it, read it, follow it.

Raw: https://raw.githubusercontent.com/awizardxch/Spellbook/main/docs/AGENT_ONBOARDING.md

The 60-second version

Enough to know whether this is for you. The repo file has the exact contract.

01โš™๏ธ

Install the daemon

Run install.sh on your machine. It provisions the daemon, two tokens (request = yours, approval = your human's), and the paper backup โ€” printed once, on paper, offline.

02๐Ÿ“ก

Point at a relay

Deploy the Chia relay (relay/) or use your operator's. It's a bearer-authed HTTPS API for your agent code โ€” coins by puzzle hash, signed-bundle broadcast. It never sees keys.

03๐Ÿง™

Request, never approve

Use the AgentClient to propose spends and surface decoded intent. Your human approves from their own tooling. You cannot approve โ€” the daemon rejects it.

Hard rules

๐Ÿ”‘

Never handle key material

No seeds, mnemonics, or private keys in chat, logs, code, or API calls. The relay rejects key material with a 400. Keys and signing stay on the daemon's machine.

๐Ÿงช

Testnet11 first

Testnet11 is the default Chia network (Solana defaults to devnet). Mainnet submission requires explicit human authorization with exact amounts โ€” it is not something you enable on your own.

The repo file is the truth.

Endpoints, auth, request/response shapes, deploy variables, and the paper-backup model โ€” all in docs/AGENT_ONBOARDING.md. If this page and that file ever disagree, the file wins.