L2BEAT

What it is

A curated metric set for Ethereum layer twos, published with the method attached. Five things are tracked across roughly a hundred projects — value secured, activity, onchain costs, liveness, and a risk assessment that ends in a Stage label — and the unusual part is not the metrics but that the definition of each one is written down, the assumptions behind it are argued in public, and the code that computes it is MIT-licensed in a repository anyone can read.

That matters more here than the category usually admits. Almost every headline number in on-chain analytics is a definition rather than a measurement, and almost every vendor ships the number without the definition. L2BEAT ships the argument: TVS is a sum of canonically bridged, externally bridged and natively minted assets, and the FAQ explains at length why that differs from a TVL figure elsewhere and why the metric was renamed when the scope changed. Activity is measured in user operations per second rather than transactions, because an L2 transaction can bundle many user actions and counting them equally would flatter whoever batches hardest. Costs are the fees the L2 itself pays Ethereum to post data, proofs and state — explicitly not the fees a user pays — split into calldata, blobs, compute and overhead.

The other half of the product is not a metric at all. Each tracked project has a hand-written assessment of its state validation, data availability, sequencing and upgradeability, and a Stage from 0 to 2 that says how far it has moved toward trust minimisation. That is an opinion, held by a named organisation that argues it in public and takes pull requests against it — which is a different kind of claim from a proprietary risk score, and the reason to use this rather than a dashboard that puts a number in a box.

Pricing

Free, in full, with no account. There is no signup, no login, no paid tier and nothing gated on the website, which is the direct consequence of how it is funded: L2BEAT describes itself as a public goods company and names Gitcoin, the Ethereum Foundation, private donors and investors, and per-project grants from companies in the space. It is a Polish limited company based in Warsaw.

Read that funding model as the trade it is. Nothing is being sold to you, so nothing is being withheld from you; and the organisation that assesses the projects is funded in part by grants from the ecosystem those projects belong to. L2BEAT is explicit about being "vocal and opinionated", and the defence it offers against the obvious conflict is the one that can actually be checked — every assessment is a file in a public repository with a history.

Data & coverage

Narrow on purpose. This is Ethereum layer twos and layer threes, and nothing else: no Solana, no Bitcoin, no multi-chain comparison of the sort the rest of this category sells. Within that scope the current summary counts 23 rollups, 4 validiums and optimiums, and 74 "others" — projects tracked because they are credibly moving toward trust minimisation rather than because they have arrived — securing $43.18 billion in total, of which rollups hold $33.82 billion.

Coverage is not uniform across the five metric families, and the difference is easy to miss. Value secured, activity and data availability cover the full roster. Costs and liveness cover fewer — 21 rollups, 3 validiums and optimiums and 51 others — because both depend on identifying the project's own Ethereum transactions, which is work done per project. If you are comparing cost per user operation across the ecosystem, you are comparing the subset that has been instrumented.

The data availability section reaches beyond Ethereum, which the scaling section does not: it tracks 5 public DA layers including Celestia, Avail and EigenDA plus 17 custom ones, with their own risk columns for the DA layer and its bridge. Value-secured series move at six-hour granularity.

There is no entity or address labelling here of any kind. What L2BEAT attributes is projects — which contracts a rollup runs, who holds its upgrade keys, which actors are permissioned — and the provenance of that is a file in the monorepo rather than a proprietary label set. It is the only product in this category whose attribution layer can be read line by line, and the only one where you can open a pull request when it is wrong.

Integrations

The interface question has an awkward answer. A documented API exists — there is a Swagger UI at api.l2beat.com/docs — but it is key-gated, and so is its own specification: requesting the OpenAPI document returns 401 with "Unauthorized. Use apiKey query parameter with valid API key." Nothing on the website, in the FAQ, in the repository README or in a search says how a key is obtained, whether it costs anything, or what the limits are.

Separately, the website's own backing endpoints answer without authentication and return the same series the charts draw. They are undocumented, appear nowhere in the published API surface, and are an implementation detail of the frontend rather than an interface anyone has promised to keep — which is exactly how a production dependency breaks quietly. If you need this data programmatically and cannot get a key, the honest options are the repository or a scraper you maintain yourself.

Limitations

• Ethereum only. L2s and L3s that settle to Ethereum. Not a multi-chain analytics platform and not trying to be one.
• The documented API is gated and undocumented in the ways that matter. No published route to a key, no published price, no published rate limits, and a specification that itself answers 401.
• The unauthenticated endpoints are not a published interface. They work today; nothing says they will work tomorrow, and nothing says what the terms of using them are.
• Costs and liveness cover about three quarters of the roster, not all of it.
• Stages and risk assessments are judgements, held by a funded organisation, updated by hand. That is the product, and it is better documented than its competitors' — but it is not a measurement, and a project's stage can change because the assessment changed.
• No alerting, no export controls, no account. There is nothing to configure and nothing to save.
• The terms of service say nothing about reusing the data. Last updated 28 October 2024, they cover prohibited conduct — illegal activity, unauthorised access, abuse, market manipulation — and contain no redistribution, attribution or commercial-use clause in either direction. The MIT licence covers the code in the repository; it is a software licence and it does not address the hosted numbers. If you intend to republish L2BEAT figures commercially, that is a question to ask rather than a permission to infer.
• The site can be discontinued without notice, which the terms reserve explicitly. For a grant-funded public good that is a real line rather than boilerplate.

Alternatives

Nothing else in this category does the same job. DefiLlama overlaps on value but measures something different and says so, and the difference between the two numbers for the same chain is a methodology gap rather than an error. A curated-metric platform with a business model will cover more chains and answer more questions and will not show you how any of it is computed. The pairing that actually works is L2BEAT for what secures a chain and what it costs the chain to be secured, and an economics-oriented metric set for what the chain earns — read side by side, with the knowledge that the second one probably reads its stage labels from the first.