Lightning Labs’ Wavelength: Bitcoin’s Third Ark for AI Agents

Lightning Labs, the company behind the software that still runs a plurality of Bitcoin’s Lightning Network, spent July 21 introducing a project that looks less like a Lightning upgrade and more like a hedge against Lightning’s own architecture. Wavelength is a self-custodial payments toolkit built on an Ark-style settlement layer, aimed at developers, AI agents, and what the company’s own announcement calls “vibe coders.” It is Lightning Labs’ first major bet outside the payment-channel model that made Lightning famous, arriving at a moment when the base network is showing real friction: a shrinking public node count, an unsolved jamming problem, and a user experience that still asks people to think about channels and liquidity before they can send money.

A company that built Lightning now builds around it

Wavelength did not appear in isolation. It is the third notable Ark-style protocol to reach the market in less than a year, after Ark Labs shipped Arkade’s public beta in October 2025 and Lightspark’s Spark network became the settlement layer quietly running behind consumer features like Cash App’s Bitcoin tools and Polymarket’s instant deposit flow. What makes Wavelength notable is who built it. Lightning Labs wrote LND, the node implementation that still runs much of the Lightning Network’s public infrastructure; co-authored much of the BOLT specification that defines how Lightning works; and built Taproot Assets, the protocol now carrying Tether’s USDT across Lightning channels. A company with that much invested in Lightning’s existing payment-channel model choosing to build a parallel, channel-free architecture is as much a signal as the product itself.

According to Lightning Labs’ announcement and product documentation, Wavelength is available today in alpha on Bitcoin’s signet and testnet networks, with mainnet access gated behind an invitation request, typical for infrastructure this early. The stated goal is simple: let a developer add Bitcoin payments to an app without running a node, managing a channel, or sourcing inbound liquidity, the three tasks that have kept Lightning largely the province of specialists eight years after the network’s mainnet launch, according to Bitcoin Magazine’s coverage of the launch.

What Wavelength actually does

Strip away the branding and Wavelength is an embedded, self-custodial Bitcoin wallet that a developer can drop into an application much like a conventional payments API. The SDK ships in several forms: a WebAssembly build that runs directly in a browser, a React Native module for mobile apps, native iOS and Android libraries, a gRPC and REST API for backend integration, and a command-line client, according to Wavelength’s own documentation. Keys are generated and held on the user’s device; Lightning Labs’ coordination service never takes custody of funds and, per the company, cannot move them unilaterally.

The wallet supports three ways of moving bitcoin. Plain on-chain transactions work as they always have. Lightning payments route through Loop, Lightning Labs’ existing submarine-swap service, which lets a Wavelength wallet send and receive across the broader Lightning Network without holding open channels of its own. Transfers between Wavelength users can also settle through the new Ark-style layer, batching many users’ transfers into shared, periodically-settled on-chain transactions instead of routing each payment through an individually managed channel. All Lightning-routed payments still rely on standard BOLT 11 invoices, so a Wavelength wallet can pay, and be paid by, any other Lightning wallet or node already on the network; Wavelength adds an on-ramp rather than a separate payment universe cut off from everyone else.

Pricing during the alpha is straightforward: Lightning Labs is charging a 1 basis point service fee on Lightning-routed transactions, plus standard on-chain network fees whenever a user settles to the base chain.

Inside the architecture: what “Ark-style” actually means

The Ark model Wavelength borrows from solves a specific problem: Lightning channels require two parties to lock funds together and stay roughly reachable to manage that channel, which does not fit a mobile app user who opens their wallet twice a month. Ark-style protocols instead let many users share a single on-chain UTXO through a structure of pre-signed, cooperatively constructed transactions. According to Bitcoin Optech’s technical overview of Ark, the basic unit of value is a VTXO, a virtual UTXO representing a user’s claim on a shared on-chain output. Periodically, users and an operator, what Ark’s own documentation calls an Ark Service Provider, jointly build a tree of transactions in a process called a round, then broadcast only the root transaction on-chain, spreading one on-chain fee across every participant in that round.

Two features keep this trust-minimized rather than fully custodial. First, every VTXO carries an absolute timelock; if the operator disappears or refuses to cooperate, a user can unilaterally broadcast their own branch of the transaction tree and reclaim their funds on-chain without needing anyone’s permission. Second, day-to-day transfers between users happen through cooperative, co-signed updates rather than the operator moving funds on its own authority. The tradeoff, which Bitcoin Optech’s documentation states plainly, is that off-chain payments still require the operator’s cooperation to finalize, and a dishonest operator colluding with a counterparty could in theory attempt to double-spend against a VTXO holder who is not paying attention. Users also have to actively refresh their VTXOs before the timelock expires, an extra piece of upkeep that channel-based Lightning does not require.

Wavelength layers its own wallet and SDK on top of this idea, offering VTXOs as a settlement option alongside on-chain and Lightning-via-Loop payments, and packaging the whole stack behind a single API so a developer never has to reason about trees, rounds, or timelocks directly.

The third Ark: how Wavelength stacks up against Arkade and Spark

Ark Labs shipped first. Arkade’s public beta went live on Bitcoin mainnet in October 2025, which the company described at the time as Bitcoin’s first major Layer 2 launch since Lightning itself. Ark Labs CEO Marco Argentieri was blunt about the state of the field when Arkade shipped. “The Bitcoin L2 landscape has been full of promises but light on shipping,” he said, adding that the release marked “the beginning of Bitcoin’s evolution as programmable money.” Ark Labs has since raised a $5.2 million seed round backed by Tether, announced in March 2026, aimed at building what it calls Arkade Assets, a multi-asset and stablecoin framework layered on the same VTXO architecture, according to CoinDesk’s report on the funding round.

Spark, built by Lightspark, took a related but distinct path: statechains, a concept Ruben Somsen first proposed in 2018, combined with FROST threshold signatures. Rather than a single operator, Spark currently runs on a small set of them (Lightspark and Flashnet, as of mid-2026) under what the project describes as a 1-of-n honesty security model, meaning the system stays safe as long as at least one operator behaves correctly. Spark is already the infrastructure quietly running underneath consumer-facing products, including Polymarket’s instant, self-custodial Bitcoin deposit flow. Lightspark CEO David Marcus has been candid for years about why his company built something other than plain non-custodial Lightning in the first place. “If you want full support for non-custodial Lightning with offline receive and want to make it economically viable, you have to accept some form of compromise on the trustlessness level of the solution,” Marcus said, according to Bitcoin.com’s report on his comments. That is the same tradeoff every Ark-style system, Wavelength included, is ultimately making.

Wavelength’s distinguishing bet is less about the settlement mechanics, which are broadly similar to its two predecessors, and more about distribution and audience. It ships as a drop-in SDK from the company that already maintains the node software much of the network runs, it bundles first-class support for machine-to-machine payments from day one, and it is aimed explicitly at fast-moving app and agent developers who were never going to run an LND node in the first place.

Built for machines: AI agents, MCP servers, and L402

Wavelength’s most forward-looking piece is its support for autonomous agents. The SDK ships with a Model Context Protocol server, the increasingly standard way AI agents connect to external tools, letting an agent request a Bitcoin payment as a native tool call rather than through a bolted-on plugin. It pairs this with L402, Lightning Labs’ existing protocol that repurposes the largely dormant HTTP 402 “Payment Required” status code so an API can demand a small Lightning payment before it responds, which an agent can pay programmatically with no account, API key, or signup flow required.

L402 itself is not new. Lightning Labs’ Michael Levin, the company’s head of product growth, introduced an open-source toolkit built around it in February 2026: seven composable skills covering node operations, remote key isolation, and scoped credentials, plus lnget, an L402-aware command-line HTTP client, and Aperture, a reverse proxy that turns any existing API into a pay-per-use endpoint. Levin’s framing at the time was that the tools let agents transact “directly on a bitcoin-native payments rail without requiring identity, API keys, or signup flows,” according to The Block’s coverage of the launch. Wavelength effectively folds that L402 toolkit and a self-custodial wallet into a single package, aimed at the same problem: agents that need to pay for API calls, data feeds, and compute in amounts far too small for a card-payment rail to bother processing.

Lightning is not the only rail chasing this use case. Coinbase introduced its own Agentic Wallets and backed the x402 payments protocol around the same time Levin’s L402 toolkit shipped, letting agents hold funds, trade, and pay for services over Coinbase’s Base network using USDC rather than bitcoin. Stripe has previewed x402 support of its own for machine payments in USDC, and CoinGecko has already turned on x402-gated API endpoints priced at a fraction of a cent per request. Google has pushed a parallel agent-payment framework, AP2, into the same conversation. The pattern across all of them is the same: none of today’s payment rails, card networks included, were built to clear a payment worth a tenth of a cent for a single API call, and Bitcoin’s Lightning-derived rails are now competing directly with stablecoin-based rails to become the default plumbing once agents start paying each other at machine speed. Whether AI agents are actually generating meaningful payment volume yet, as opposed to mostly demo traffic ahead of real adoption, remains an open and contested question across every one of these rails.

Practical use cases beyond AI agents

While AI agents get top billing in Lightning Labs’ own marketing, Wavelength’s SDK shape targets a broader set of problems that have dogged Bitcoin payment integration for years. Content monetization is an obvious one: a publisher could gate an article, a video, or an API response behind a fraction-of-a-cent Lightning payment instead of a subscription, the exact problem L402 was originally built for and that Aperture already supports as a reverse proxy. In-game economies are another: a mobile game could let players buy or sell in-game items for small bitcoin amounts without the studio running Lightning infrastructure itself, the kind of integration HOGE Wire’s look at why Web3 gaming skipped Lightning found studios avoided specifically because of channel-management overhead and Bitcoin’s lack of general smart-contract support. Machine-to-machine commerce beyond chatbots fits too: IoT devices, data marketplaces, and automated trading systems that need to exchange small, frequent payments without a human approving each one.

None of these use cases are unique to Wavelength; they are the same ones Arkade, Spark, the original L402 toolkit, and years of Lightning-native startups have chased with varying success. What Wavelength adds is a single SDK that bundles on-chain, Lightning, and Ark-style settlement together, letting a developer pick a mode without necessarily needing to understand the tradeoffs between them upfront, plus a distribution channel through Lightning Labs’ existing relationships with wallets, exchanges, and infrastructure providers that smaller Ark-style startups do not have.

Why Lightning Labs needed its own answer to the last-mile problem

None of this happened in a vacuum. Base-layer Lightning has structural friction that eight years of protocol work has narrowed but not eliminated. A new node cannot receive payments until it has inbound liquidity, typically purchased from a liquidity provider or built up by first spending outbound capacity, which is a strange first experience for anyone arriving from a bank app or a custodial exchange. Running a node well enough to route payments reliably takes ongoing attention: channel rebalancing, backup management, and monitoring for force-closes. Public capacity has also been shrinking on some measures even as usage grows, a tension covered in more depth, including the still-unsolved channel-jamming denial-of-service problem and the concentration of node hosting among a small number of cloud providers, in HOGE Wire’s separate look at Lightning’s record capacity and unsolved jamming risk. That kind of friction is exactly what pushes ordinary users toward custodial wallets and pushes companies building consumer apps to look for something that behaves more like a normal API, which is the gap Wavelength, Arkade, and Spark are all trying to fill from slightly different angles.

The state of the base Lightning Network today

Wavelength and its Ark-style peers are explicitly built as alternatives to opening and managing individual payment channels, but they still ultimately rely on Bitcoin’s base layer for security and, in Wavelength’s and Spark’s cases, on Lightning itself for routing payments to wallets and nodes outside their own systems. It is worth taking stock of what that base layer looks like heading into the back half of 2026.

Public node count stood at 17,438 as of that same May 2026 reading, down about 16% from the network’s 2022 peak of roughly 20,700, according to Spark’s state-of-the-network research. Monthly Lightning payment volume reached roughly $1.17 billion in November 2025, the most recent figure in that same research, up sharply from the prior year and continuing a multi-year growth trend even as raw capacity and node counts have pulled back from their late-2025 highs. Bitcoin itself was trading near $65,800 on July 22, 2026, according to CoinDesk. The gap between rising payment volume and falling headline network-size metrics is the core tension every Ark-style project, Wavelength included, is implicitly betting on: that most future Lightning-adjacent activity will run through a smaller number of well-capitalized operators, wallets, and settlement layers rather than a large number of individually managed channels.

Where Wavelength fits in Bitcoin’s dollar-rails race

Wavelength arrives alongside, not instead of, an increasingly crowded field of projects trying to move dollars over Bitcoin rails. Tether’s USDT has been live on Bitcoin and Lightning since early 2026 through Taproot Assets, the Lightning Labs-built protocol that lets a stablecoin ride Lightning’s channel graph while bitcoin remains the routing and settlement asset. A second, competing path exists too: RGB, a client-side-validation protocol with roots tracing back to Peter Todd’s 2014 single-use-seals concept, is being brought to Bitcoin’s mainnet commercially by UTEXO, a startup that raised $7.5 million in seed funding led largely by Tether in March 2026. The goal, according to UTEXO co-founder Viktor Ihnatiuk, is to let “USDT move on Bitcoin the way money is supposed to move: instantly, privately, with no surprises on costs,” as he told Bitcoin Magazine.

Wavelength’s public documentation frames its own roadmap around bitcoin first, but the SDK’s multi-asset ambitions and its explicit machine-payment focus put it on a collision course with both Taproot Assets and RGB as candidate rails for stablecoin-denominated agent payments, the use case every one of these projects keeps returning to. An AI agent paying for API calls a thousand times a day cares less about which settlement protocol it is using than about fees, finality time, and whether the wallet holding its funds has to also hold volatile bitcoin. That is a real design question none of Ark, RGB, or Taproot Assets has fully answered yet.

The regulatory backdrop

Any project touching agent-to-agent or app-to-app payments in 2026 is building against an unusually unfinished regulatory backdrop, at least in the United States. The GENIUS Act, which established a federal framework for payment stablecoins, gave regulators exactly one year from its July 18, 2025 signing to finalize implementing rules, and that deadline passed on July 18, 2026 with the Office of the Comptroller of the Currency, the FDIC, the National Credit Union Administration, and the Federal Reserve all still working from proposed rather than final rules, according to reporting on the missed deadline. The law’s substantive requirements still take effect automatically eighteen months after enactment regardless of whether the rulemaking is finished, so the rules will eventually bind, just not on the original schedule.

None of that uncertainty is specific to Wavelength, which for now moves bitcoin rather than a stablecoin. But the toolkit is explicitly designed to add multi-asset support later, at exactly the moment when the compliance path for the stablecoins most likely to ride on top of it is still being written. Any developer building an agent-payment product on Wavelength, Taproot Assets, or RGB today is implicitly betting the rules land somewhere workable, without yet knowing the details.

Risks and open questions

Wavelength is alpha software with mainnet access still gated behind an invitation list, and several of its open questions are the same ones that apply to Ark and Spark generally. The coordination service, whatever Lightning Labs ultimately calls it in production, has to stay honest and available; a user’s unilateral exit right protects their funds if the operator disappears, but exercising that right means falling back to an on-chain transaction, with on-chain fees and confirmation times, not the instant experience the product is designed around. Off-chain transfers between users still require the operator’s cooperation to finalize, the same counterparty-risk tradeoff Bitcoin Optech flags in its technical description of Ark. VTXOs also need to be refreshed before their timelocks expire, an obligation that shifts some of the operational burden Wavelength is trying to remove from end users onto whatever wallet or service is managing that user’s keys on their behalf.

There is also a fragmentation question hanging over the category as a whole. Bitcoin now has three prominent Ark-style systems, Arkade, Spark, and Wavelength, built by three different companies, each with its own operator set and, in Spark’s and Wavelength’s cases, its own trust assumptions about who is allowed to coordinate. A payment sent from a Wavelength wallet to a Spark wallet still has to route over Lightning proper or settle on-chain; the three systems are not natively interoperable with each other, only with the base layer they all sit on top of. Bitcoin’s broader history with self-custody products argues for caution here too: key management remains one of the most common ways users lose funds, whether through a compromised device, a phishing link, or a transaction signed blind without being fully understood, a problem HOGE Wire has documented at length in the context of multisig wallet security and blind signing. Wavelength’s on-device key model avoids custodial risk in the traditional sense, but it inherits every device-security and phishing risk that comes with self-custody, now extended to AI agents that may sign and send payments with far less human oversight per transaction than a person tapping confirm on a phone.

Software this new and this security-sensitive is also exactly the kind of target the industry’s independent security researchers exist to stress-test before real money moves through it at scale, work HOGE Wire covered in its look at the whitehat bug bounty workforce that increasingly audits new Bitcoin and Lightning infrastructure before, not just after, it reaches mainnet.

Industry reaction

Reaction to Wavelength’s launch across Bitcoin developer circles has been a mix of genuine interest and a certain amount of category fatigue. One widely shared social post summarized the tone concisely, framing Wavelength as “a 3rd Ark” and Lightning Labs’ answer to “lightning’s last mile problem,” language close enough to Lightning Labs’ own positioning that it reads as much as description as commentary. That a company as central to Lightning’s existing infrastructure as Lightning Labs felt compelled to build its own Ark-style alternative, rather than simply improve channel-based Lightning further, is itself being read by parts of the developer community as a quiet acknowledgment that the payment-channel model has limits for consumer and machine-payment use cases that better liquidity tooling alone will not fix.

Exchanges and wallets, meanwhile, are the distribution layer that will ultimately decide whether any of this matters outside developer circles. Lightning withdrawals and deposits are already live at every major exchange, a landscape HOGE Wire mapped out in its comparison of withdrawal experience across Coinbase, Binance, Kraken, and OKX. None of the big four has said anything publicly yet about integrating Wavelength specifically, and with mainnet access still invitation-only, that is not a decision any of them needs to make in the near term.

What comes next

Lightning Labs has not published a public timeline for moving Wavelength from invitation-only mainnet access to general availability. The company’s own materials describe the current release as an alpha, and the emphasis in its documentation on developer experience, an installable SDK, a hosted demo, and multiple language and platform bindings, suggests the near-term goal is attracting builders rather than end users directly. Given the path Arkade took from public beta to broader integration work, and the way Spark grew from a Lightspark-only project into shared infrastructure now used by outside apps like Polymarket, the more interesting test for Wavelength will not be its alpha metrics but whether unaffiliated developers choose to build on it once mainnet opens more broadly, particularly for the AI-agent payment use case Lightning Labs is most visibly chasing.

For now, Wavelength is best understood less as a finished product and more as a statement of intent from the company most responsible for Lightning as it exists today: that the next several years of Bitcoin payments infrastructure may look less like everyone running their own node, and more like a small number of specialized settlement layers, Arkade, Spark, Wavelength, and whatever comes next, competing for the same developers, the same agents, and ultimately the same users.

Frequently Asked Questions

What is Wavelength and who built it?

Wavelength is a self-custodial Bitcoin payments SDK built by Lightning Labs, the company behind the LND node software that runs much of the Lightning Network and the Taproot Assets protocol that carries USDT over Lightning. Announced on July 21, 2026, Wavelength lets developers add Bitcoin and Lightning payments to an app through an embedded wallet and a simple API, without the app or its users needing to run a Lightning node, open payment channels, or manage inbound liquidity themselves. It is currently in alpha, live on Bitcoin’s signet and testnet networks, with mainnet access available by invitation.

How is Wavelength different from the regular Lightning Network?

The Lightning Network moves bitcoin through payment channels that two parties open, fund, and manage directly. Wavelength instead uses an Ark-style settlement layer, where many users share a single on-chain UTXO through a tree of pre-signed transactions coordinated by an operator, removing the need for individual channel management. Wavelength wallets can still send and receive standard Lightning payments through Lightning Labs’ Loop swap service, so they remain compatible with the wider Lightning Network. The tradeoff is that day-to-day transfers rely on a coordinating operator, with a fallback on-chain exit if that operator ever stops cooperating.

Is Wavelength the same thing as Ark or Spark?

No, though all three are built on a related idea. Ark Labs’ Arkade and Lightspark’s Spark are the two earlier Ark-style systems to reach Bitcoin, launched in October 2025 and earlier in 2026 respectively. Wavelength is Lightning Labs’ own implementation of the same broad concept: users sharing on-chain UTXOs through VTXOs and periodic settlement rounds coordinated by an operator. Each project runs its own independent operator infrastructure and SDK, and the three systems are not natively interoperable with one another; a payment between a Wavelength wallet and a Spark wallet still has to route over base Lightning or settle on-chain.

Can AI agents actually use Wavelength to make payments on their own?

Yes, that is one of Wavelength’s primary design targets. The SDK ships with a Model Context Protocol server, letting AI agents built on MCP-compatible frameworks request Bitcoin payments as a native tool call, and it integrates Lightning Labs’ existing L402 protocol, which lets software pay for an API call or data feed automatically using the HTTP 402 status code, with no human approval, account, or API key required for each transaction. The goal is letting autonomous agents pay for compute, data, and services in amounts far smaller than traditional card-payment rails can economically process.

Is Wavelength available to use right now?

Partially. Wavelength’s SDK and alpha network are live today on Bitcoin’s signet and testnet, which use test bitcoin with no real monetary value, so developers can build and experiment with it immediately. Mainnet access, where real bitcoin would move through the system, is currently invitation-only, and Lightning Labs has not published a public date for opening it more broadly. Until then, Wavelength should be treated as early-stage infrastructure rather than a finished consumer product.

Written by the HOGE Wire Bitcoin desk.