Quickstart — connect an agent to Bitcoin
Four pathways to plug an agent into Bitcoin, from a hosted service in minutes to your own full node.
In brief. There’s no single way to plug an agent into Bitcoin — there’s a ladder, and which rung is right comes down to one question: how much infrastructure are you willing to run? This guide lays out four pathways — a custodial service (live in minutes), self-custody with no node, your own light-client node, and the full sovereign stack — and what each one trades away on custody, censorship-resistance, and effort. The rule of thumb is simple: pick the lightest pathway that meets your needs, and move down the ladder later if you need to. Every pathway ends in the same place — an agent that can pay, get paid, and reach the marketplace.
Which pathway?
Start at the top and stop at your first yes. Each step down the list trades a little more setup for a little more sovereignty.
- Don’t need to hold your own keys? → A · Custodial service. The fastest path: a hosted wallet or account holds the keys and runs the node for you.
- Want to hold your own funds, but not run a node? → B · Self-custody, no node. A headless non-custodial wallet, or a connector that drives an external wallet without touching its keys.
- Willing to run a node, but not store the whole blockchain? → C · Light-client node. Your own Lightning node on a light client — real channels, no 600-gigabyte download.
- Want to validate everything yourself? → D · Full sovereign stack. Your own Bitcoin full node underneath your Lightning node. Trust nothing.
- Handling real money? → keep your pathway and add the remote-signer hardening (below) on top of C or D.
The four pathways at a glance
| Pathway | What you run | Custody | What you trust | Time · effort | Sovereignty |
|---|---|---|---|---|---|
| A · Custodial service | nothing — an account / API | the operator holds the keys | the operator (can KYC, freeze, censor) | minutes · none | low |
| B · Self-custody, no node | a wallet or a connector | you hold the keys | the wallet / connector infra; the mint, for ecash | minutes–hours · low | medium |
| C · Light-client node | your own Lightning node on Neutrino | you hold the keys, your node | block headers (light-client level) + peer uptime | hours · medium | high |
| D · Full sovereign stack | a Bitcoin full node + your Lightning node | you hold the keys, full validation | nothing — you verify it yourself | days (incl. sync) · high | maximum |
| ＋ Remote signer (adds to C or D) | a watch-only node + an isolated signer | keys live off the agent’s host | the signer device + the limits you set | extra setup | hardens C/D |
The further down you go, the less you have to trust anyone else — and the more you run yourself. Most agents should start at the lightest rung that meets their needs.
A · Custodial service — fastest
What you run: nothing. A hosted service holds the keys and operates the node; your agent calls an API or signs into an account.
The trade-off: you’re live in minutes with zero infrastructure, but the operator can KYC you, freeze funds, or be compelled to censor. Fine for small operational balances and getting moving; not where you park a treasury.
In our stack: most of this lane lives in the Market section rather than the toolbox — see the Exchange venues and the Services directory for hosted, account-based options. On the bearer-token side, Minibits/Ippon spins up a custodial Cashu wallet with a single call (keep it small and disposable).
Start here if you just need your agent paying for an API today and don’t want to manage anything.
B · Self-custody, no node
What you run: a wallet you control, or a connector — but no node of your own.
The trade-off: you hold your own keys without standing up infrastructure. You’re still depending on the wallet or connector’s backend (and, for ecash, on a mint), so it’s not full self-sovereignty — but your funds aren’t sitting with a custodian.
In our stack: Xverse Agent Wallet is a headless, non-custodial wallet whose keys stay encrypted on-device and which answers HTTP-402 payment requests itself. Alby + Nostr Wallet Connect hands your agent a scoped, budgeted, revocable connection to a wallet hub without ever exposing the keys, and the Model Context Protocol carries those payment tools into agent frameworks. For bearer cash, Cashu gives instant, private transfers — with the honest caveat that the mint is a custodial trust point, so spread balances across mints (see how to vet a mint). Your agent pays over L402 and LNURL.
Start here if you want real self-custody quickly and are comfortable trusting a wallet or connector for the plumbing.
C · Light-client node
What you run: your own Lightning node — lnd on a Neutrino light client instead of a full chain copy. You get real channels and route payments without downloading or storing the entire blockchain. (Neutrino is an lnd capability; Core Lightning always wants a full bitcoind, so a CLN node lands in Pathway D.)
The trade-off: near-sovereign, with a fraction of the storage and sync time. The compromise is that a light client trusts block headers rather than fully validating every transaction itself — a real but well-understood reduction, not custodial trust.
In our stack: Lightning Agent Tools (Lightning Labs’ agent toolkit) runs natively in Neutrino mode; LNbits gives you programmable wallets and accounts on top of your node; Loop moves balance between on-chain and Lightning; Taproot Assets adds issued-asset rails; and Clink / Lightning.Pub is purpose-built to run an LND node in Neutrino mode on minimal hardware. Each of these confirms Neutrino support in its own docs — no full chain copy required.
Start here if you want your own node and channels but don’t want to run a full Bitcoin node yet.
D · Full sovereign stack
What you run: a Bitcoin Core full node underneath your Lightning node — the whole chain, validated by you.
The trade-off: maximum sovereignty and censorship-resistance — you trust nothing and verify everything — at the cost of the most storage, sync time, and ongoing operations.
The key thing to know: the toolkits are the same as Pathway C. Lightning Agent Tools, LNbits, Loop, Taproot Assets — they all run identically on a full node. The pathway is set by your node backend (Neutrino vs. full Core), not by the tools. That means you can start light on C and graduate to D later without changing your stack.
Want your agent to sell, not just buy? Everything above gets it paying; this same sovereign backend can also serve a priced tool over a pubkey — a CVM server with no inbound ports — instead of exposing an HTTP endpoint. See The Stack §4.
Start here if censorship-resistance is the whole point — you’re settling meaningful value and want to depend on no one.
Protect the funds: the remote signer
This isn’t a fifth pathway — it’s a hardening layer you add on top of C or D when an agent is handling real money. In a split-node setup, the agent operates a watch-only node that can initiate payments, while the private keys sit on a separate, isolated remote signer. You then issue scoped credentials (macaroons) that cap what the agent can do — pay-only, with a daily limit, say 5,000 sats. A compromised agent host can’t drain the treasury, because it never held the keys.
Lightning Agent Tools is built for exactly this (watch-only node plus a separate signer), and a scoped NWC connection gives a lighter version of the same discipline for Pathway B.
Reaching the marketplace
Whichever pathway you pick, the destination is the same: an agent that holds value and can transact for itself. Once it can pay an invoice and answer one, it can discover and buy services in the live marketplace — agent-readable, with probed liveness and a cross-provider price index. That’s the whole point of standing any of this up.