AMBOSS ⚡

Introducing the official Amboss TypeScript SDK.

Send and receive money in USDT, USDC, and Bitcoin over Lightning, straight from your backend. Sub-second settlement, no chargebacks, 24/7.

Here is what it changes for your business. 🧵

In June 2022 the average public Lightning channel held 0.048 BTC. On 1 November 2026 it held 0.097 BTC, about twice as much.

There are fewer public channels than in 2022, but each one is larger. A larger channel can carry a larger payment along a single route, which matters once real business amounts start to move.

The figure is public capacity divided by active channels, from the Amboss Space daily series.

0.048 BTC then, 0.097 BTC now.

Node operators bought 17.38 BTC of Lightning liquidity on Magma in October. In October 2025 the figure was 10.48 BTC. That is 66% more in a year.

The chart shows every month in between, so you can see the path, not just the two end points.

Magma is the Amboss marketplace for inbound liquidity: a seller opens a channel to your node for a fee agreed up front. Across all orders so far, the average price has been 4.42% a year.

Want to receive payments on your own node without asking peers one by one? Magma is built for that.

Picture a merchant taking $1 million a month in card payments with a 10% rolling reserve held for 180 days. Once the reserve fills, about $600,000 of its own revenue is held back at any time.

These numbers are illustrative, not from any one processor. The point is the arithmetic: while a rolling reserve is held, the money is earned, yet it cannot be spent.

Bitcoin and stablecoin payments through Amboss settle to your own wallet in seconds, with no rolling reserve. With optional auto-conversion, the balance can stay in US-dollar stablecoins.

If six months of reserves came back to your balance sheet, where would your team put them first?

Give a Lightning node a fixed budget. Which channels should it open so that the most payments can flow through it? Our research model, MPFlow, has made 4,640 of those channel-open decisions in production, placing 267.3 BTC across 30 managed nodes.

MPFlow treats the question as a machine learning problem. It learns which new channels raise max-flow, a measure of how much value can move between two points in the network.

In training, the biggest hubs were taken out, so the model had to learn good placement, not just "connect to the hubs". It now powers peer recommendations at Amboss.

Better placement means more payments find a route the first time.