Artemis

What it is

Artemis computes a fixed library of fundamental metrics — fees, revenue, daily active users, TVL, stablecoin supply and transfer volume — the same way across every chain and protocol it covers, so two chains can sit in one table without you reconciling two definitions. You do not write queries against raw chain data. You take the definitions Artemis has already fixed, which is the trade this corner of on-chain analytics asks you to make.

The company was artemis.xyz and is now artemis.ai, and the rebrand moved more than the domain. The front door is the Analyst, a chat that answers a question with research, SQL and charts drawn from that library, and the plans are sold in queries rather than seats. The Terminal, the dashboards, the Excel and Google Sheets add-in and the REST API all still run over the same metrics; the Analyst is a new way in, not a new dataset.

Crypto and public equities share one namespace, so a single request can return Bitcoin, Hyperliquid and a listed exchange side by side. That is unusual here, and it is also the direction the product is moving in — see the deprecations below.

Pricing

Four prices, two products, one page. The Analyst runs from a free Intern tier through Investor at 100 USD a month to Managing Director, quoted as Custom on the pricing page and as starting at 50,000 USD a year in the documentation. The Plugin — the spreadsheet add-in — is 500 USD a month per user and is bought separately.

The unit is a query, and the vendor defines it tightly: one question to the Analyst including the research, SQL and charts it produces, with follow-ups in the same thread counting again. Allowances reset monthly.

What the free tier actually gives you depends on which vendor page you read, and both were open on the same day. The documentation says 20 Analyst queries a month, one dashboard and three charts. The pricing page says "limited queries, daily rate limits" and prints no number. The two pages also disagree about whether the spreadsheet add-in is included at 100 USD a month — the pricing page lists it there with a 50,000-call allowance, the documentation's Investor card does not mention it at all.

Programmatic access is the expensive part. API keys appear in Settings only on a plan that includes them, and the documentation is blunt that if no key is listed, your plan does not have one. MCP and the Snowflake data share are Managing Director only.

Data & coverage

The chains documentation puts coverage at 112 chains and names the core set — Ethereum, Solana, Bitcoin, Base, Arbitrum, BNB Chain, Tron, Optimism, Polygon, Avalanche, Sui, Aptos, Near, TON, Hyperliquid, Linea, XRP, Cardano, Unichain, Mantle. On top of chains sit application sectors with their own metric families — lending deposits and borrow APYs, spot DEX volume and fees, perpetual volume, open interest, funding rates and order-book depth, stablecoin supply and transfers, and event-contract venues. Equities coverage is quoted at 30,000-plus tickers and 600-plus protocols.

Artemis publishes its definitions, which is the thing worth paying for and the thing most of this category will not do. Active addresses is the usual trap — senders, receivers or either — and here it is two named metrics rather than one ambiguous one. CHAIN_DAU is addresses that send a transaction on the chain. DAU is addresses using a protocol's products. The documentation shows the two agreeing exactly on Ethereum, Tron and Solana for 1 August 2026, and it also shows where the sibling pair diverges — FEES folds application-level fees on top of raw gas, CHAIN_FEES does not.

Aggregation is stated per metric, and the docs say plainly never to sum open interest. The dimensional cuts come with a warning attached too: splitting Ethereum's DAU by category produced twenty categories summing to 605,380 against an undifferentiated total of 439,738 for the same day, because an address active in two categories is counted in both. The vendor's own instruction is to treat that split as directional, not additive.

Granularity is daily and there is no finer setting. Dates are calendar days, roll-ups go to week, month, quarter or year, and the current day arrives with a null value until it closes.

Integrations

The REST API is three endpoints — list assets, list the metrics an asset supports, fetch the numbers — and one of them serves every time series in the product. The spreadsheet add-in covers Excel and Google Sheets with an =ART() formula that refreshes on pull; its Google listing was updated on 20 August 2026, is published by Artemis Analytics and reports 4,000-plus installs, which is the only install figure the vendor exposes anywhere.

The add-in is not only an Artemis client, and that is easy to miss. Its own listing names four supported data providers — CoinGecko, Artemis, Token Terminal and Dune — and the last two work by pasting in your API key, not by buying anything more from Artemis. So the same =ART() surface will pull Token Terminal metrics and the results of your own Dune queries into the same sheet, on the contracts you already hold with them. That makes the plugin worth more to somebody who already pays those two than the price of the Artemis data alone suggests — and it means a sheet built on it can break for reasons that have nothing to do with Artemis.

Snowflake customers get a data share, which the vendor describes as a separate delivery surface on its own update schedule rather than a mirror of the API. An MCP server is offered on the top tier.

One gap worth knowing before you plan around it. The documentation index describes access "via Terminal, Sheets, REST API, Python SDK, and Snowflake Data Share", but there is no SDK page anywhere in that index. The SDK exists and is simply unlisted: artemis on PyPI, which describes itself there as "the official Python library for the artemis API", Apache-2.0, machine-generated, last released as 0.5.0 in May 2026 and still sending readers to the pre-rebrand app.artemis.xyz/docs. Read the three HTTP endpoints rather than the client, because the client is not something the vendor documents or appears to be keeping current.

Limitations

Coverage is actively pruned, and recently. On 27 May 2026 Artemis turned off the data pipelines for 122 chains and protocols "where customer demand had significantly declined" — Berachain, Blast, Celo, Blur, CoreDAO, Axelar, Aevo and Astar among them — and stopped supporting them entirely across Snowflake, API, Sheets and Terminal on 26 June 2026. Existing history stays where it exists. Nothing new is computed for any of them.

The application layer thinned out two weeks later. On 12 June 2026 a list of application- and category-level activity metrics was removed, explicitly "as part of a large migration to move more towards equities": GAS_FEES_USD, CONTRACT_COUNT, REAL_USERS, the L7/L30/L90 DAU averages and the new-versus-returning user split. DAU and DAILY_TXNS survive for chains and protocols but not for long-tail applications and categories. If your use of this product is per-application user behaviour rather than chain and protocol fundamentals, check the deprecation list before you build on it.

A field name can change meaning underneath you. In July 2026 stablecoin transfer volume moved from an adjusted figure — deduplicated, with intra-exchange transfers and MEV removed — to gross transfers minus mints and burns. Volume reads higher as a result, and the vendor states that API field names are unchanged, so the same field now returns a different quantity. The filtered ARTEMIS_* and P2P_* flavours are being sunset. A chart that crosses that date is two measurements.

Errors arrive as HTTP 200. When a metric is not available for an asset, or a cut does not apply, the response carries a string where the array would be. Every symbol you ask for comes back, including one that does not exist, so a naive client sees a full response and no failure.

The public changelog is thin and says so. Between 30 July 2025 and 27 May 2026 it records nothing, and the page states it is not exhaustive and directs anyone with an integration to email support. That is a poor basis for noticing a definition change in time.

There is no cheap way to automate this. API access starts at the 500-dollar-a-month spreadsheet seat; MCP and Snowflake start at Managing Director. For a single developer pulling a handful of series, the price of this dataset is the price of a desk, not of an API key.

Nobody on this site has held a paid Artemis account, so everything above is read from the vendor's own documentation and pricing pages on 19 September 2026, not from use.

Alternatives

Token Terminal covers the same fundamentals with the same insistence on one definition everywhere, framed as an income statement rather than as a metric catalogue, and its API is quoted rather than priced — so the two are alike in the thing that hurts, which is what automation costs.

DefiLlama is the free answer for TVL, fees and revenue, with the definition written down in public, and the honest trade is that its adapters are usually written by the protocol being measured. If the subject is Ethereum layer twos specifically, L2BEAT and growthepie are both free and openly licensed — and on growthepie the chains are the customers, which is the same kind of thing to know as who pays for a listing here.

CoinGecko API is the cheap way to get prices and market data alongside none of this; it is a different dataset, not a cheaper one.