Why bolthub Is Lightning-Only (and How We Make Running a Node Painless)
We built an x402 rail. We are removing it.
A while back we added a second payment rail to bolthub: x402, the protocol for settling USDC on-chain. We shipped it behind a flag, kept it dark, and set ourselves a rule: revisit the decision with real data instead of vibes. We have now done that, and we are reverting to a Lightning-only platform.
This post is the honest why. It is a strategy call, not a technical failure. x402 works. It is just not the rail for what we are building.
The thesis bolthub is built on
bolthub is a bet that machine payments are tiny, constant, and instant. An agent calls a tool, pays a fraction of a cent, gets a result, and does it again a thousand times an hour. Card networks cannot price that low. A sat is about $0.001, roughly 500× below the card floor, and it settles in under a second.
Lightning is built for exactly that shape of payment. The companion post walks through the numbers, and the short version is this: on-chain gas is a fixed cost per settlement that does not shrink as the payment shrinks, while a Lightning routing fee scales down with the payment. For sub-cent, high-frequency traffic, Lightning is cheaper and faster, and the gap grows the smaller each payment gets.
If you believe, as we do, that the machine economy pays in fractions of a cent, then the rail question mostly answers itself.
Why not just support both?
The obvious hedge is to run both rails and let customers choose. We tried that framing, and we do not think it serves users here.
- •Optionality is not free. Two rails means two settlement paths, two liquidity stories, two sets of edge cases, and two things to secure and explain. That surface has a cost, and it is paid in focus.
- •A muddy product is a worse product. "Settle on any rail" sounds flexible. In practice it asks every user to make a decision we are better positioned to make for them.
- •The small end is where we win. For the sub-cent payments we care about, one rail is clearly better. Splitting attention across two would make us worse at the thing we are actually good at.
So we are choosing, on purpose. bolthub is Lightning-only, and that is a feature, not a limitation.
The one honest objection: you need a node
Being honest about Lightning means being honest about its cost. To receive Lightning payments reliably you need a node with inbound liquidity, and it has to stay online. This is the thing x402 lets you skip, because there you just settle to an address.
A phone wallet does not solve it. Your buyers are agents, and they pay you around the clock, including while you sleep. The wallet that issues your invoices has to be up 24/7. That rules out a wallet on a device you carry around and turn off.
This is the real tax on the Lightning side. We are not going to pretend it away. Instead we spend our effort making it small.
How Node Launcher removes the tax
Node Launcher is our answer to the node requirement. It deploys your own non-custodial LND node on a VPS you control, in minutes, from about $3.50/month. You hold the keys. bolthub never touches the funds and cannot move them.
- •Non-custodial by construction. Your seed is generated in your browser and never leaves it. bolthub holds no credential for the machine and adds no fee on top of the VPS bill.
- •Guided inbound liquidity. You order a channel and it opens automatically, brokered across providers with a fallback. That gives you the inbound capacity you need to actually receive payments.
- •Monitoring and alerts. It checks your node every few minutes and emails you before capacity runs low, so you are not the last to know.
- •Backup and one-click recovery. Channel state is backed up encrypted, in a form bolthub cannot read, and you can restore from your seed.
If you already run Lightning, you do not need any of this. Connect your existing LND or LNbits, or any always-on NWC wallet like Alby Hub. Node Launcher is for everyone who is not already set up, which is most people. It turns "run a Lightning node" from a project into a few minutes of setup.
Where this leaves stablecoins
We are not against stablecoins, and this is not a door slammed shut. L402 is a rail-agnostic primitive at the tool-call boundary, and if the machine economy shifts toward larger, on-chain-shaped payments, we can revisit. But we will not ship a rail we cannot make cheaper and faster than Lightning for the payments our users actually make. Today, for sub-cent machine payments, that rail is Lightning.
So bolthub is deliberately, strategically Lightning. If you want to charge agents per call and keep custody of every sat, that is exactly what we built.