Published pricing in agentic data engineering

1. The census

Rows are ordered by classification, then as listed. Data Workers sits in sequence with everyone else, fetched the same way on the same day.

Every figure in the table is quoted from the URL in the same row, as it read on 10 September 2026. Nothing in this table is a negotiated deal value, an estimate or a figure from a third party.

2. Findings

Each finding is one claim, and each one is read off the rows above rather than asserted.

1. 1.6 of 21 vendors publish a paid rate on their own site, checked on 10 September 2026. The rate-card rows: Alkera, Altimate AI, Ardent AI, Recce, Wren AI and Data Workers. Every other row's figure cell reads "does not publish".
2. 2.5 of 21 vendors have no pricing page at all. Upriver, Actioneer, Velum Labs and Mica AI each returned HTTP 404 at /pricing on the fetch date. Google publishes no pricing page for the BigQuery data engineering agent itself.
3. 3.The modal outcome is a pricing page with no money on it: 10 of the 16 vendors that have a pricing page show no figure. 6 name a billing unit without a rate and 4 name neither. A published unit with no rate is the single largest classification tied with the rate cards.
4. 4.0 of the 3 platform vendors showed a dollar rate on the pricing page fetched, against 3 of the 5 direct peers. Databricks defines DBU, DSU, GB and HOUR and Snowflake names credits and per-TB storage, but on both pages the figures sit behind a region selector, a calculator or a linked consumption table rather than on the page. Among the direct peers, Alkera and Altimate AI show tier prices in the markup.
5. 5.5 of the 6 published rate cards attach a consumption meter to the subscription figure. Altimate AI meters tokens, Ardent AI meters CU-hours and GB-months, Recce meters agent reviews, Wren AI meters credits. Alkera's page showed tier prices with no separate meter. Data Workers publishes no usage meter.
6. 6.One vendor in the census publishes a dollar rate for model tokens on its own pricing page. Altimate AI: $5 per 1M tokens, top-up rate, on https://altimate.ai/pricing.

3. Where Data Workers sits

Data Workers is one of the 6 vendors in this 21-vendor peer set that published a paid rate on its own site on 10 September 2026, and the only one of the 6 whose published rate carries no usage meter.

When not to use Data Workers. If you need a SOC 2 Type II report before any pilot, we are not a fit yet: SOC 2 is not yet certified and a Type I audit is planned with timing to be confirmed, as stated on our security page. If you want a free hosted tier you can click into before lunch, several vendors in the table above offer one and we do not. If you want alerting rather than agents that propose and apply changes under a named human’s approval, an observability tool is the better purchase.

4. Method

The frame. The 21 vendors are exactly the vendors of The Agentic Data Engineering Index, Q4 2026. No vendor was added or dropped to make the count land anywhere. Where this census and the Index’s own publishes.price field disagree about a vendor, this page is the current record: the Index field was recorded on 9 September 2026 against a looser test, and every row here was re-fetched on 10 September 2026 against the four-class definition below.

What was counted. The vendor’s own website, and nothing else. One pricing URL per vendor plus at most one supporting fetch, a hard cap of two fetches per vendor, all on 10 September 2026. The URL and the date are recorded on every row.

What “published” means here. A vendor is recorded as publishing a rate card only when at least one paid tier shows a money figure a buyer can read without contacting sales. A free tier alone does not qualify. A named billing unit with no rate attached does not qualify, and gets its own class.

The four classes.

• •Published rate card. At least one paid tier shows a number.
• •Published unit but no rate. The page names the metering unit, a credit, a DBU, a seat, a data asset, but shows no money figure.
• •Contact sales only. A pricing page exists and shows neither a paid rate nor a metering unit.
• •No pricing page. The pricing URL returns 404, redirects away, or the site has no pricing page. A 404 is a finding and is recorded as one.

What was excluded. AWS Marketplace and other reseller listings, analyst reports, third-party blogs, pricing quoted in sales calls, and any figure not visible on the URL in the row. Where no price was visible the cell reads “does not publish”, which is a statement about the page and not about the vendor.

How to contest a row. Email hello@dataworkers.io with the vendor, the URL and what the page shows. If the row is wrong we correct it and say what changed, and upheld corrections are credited by name in the next edition. That applies to our own row as much as anyone else’s.

How the page and the dataset stay in step. One script writes both the table you are reading and the JSON linked in section 8, from a single set of rows. Neither is hand-edited, so they cannot disagree.

5. How this differs from our August report

In August 2026 we published a market pricing report that read 52 pricing pages on 9 August 2026, counted 19 vendors that had invented their own billing unit, 10 that took a margin on model tokens, and catalog contracts on AWS Marketplace listed between $48,000 and $100,000 a year. This page is narrower and stricter: 21 agentic data engineering vendors instead of 52 across the wider data tooling market, one row per vendor, every row dated and sourced to a single URL, refreshed quarterly.

Where the two look like they disagree, they are describing different surfaces. The August report’s $48,000 to $100,000 range comes from AWS Marketplace contract listings, including Atlan at $100,000 and Monte Carlo at $50,000 for a 12-month contract. This census excludes marketplaces, so Atlan appears here as a contact form and Monte Carlo as credits with no cost per credit. Both readings are accurate on their own surface, and a buyer should know that a vendor publishing nothing on its own site may still have a readable contract price one click away on a marketplace.

Two things the August report claimed are not re-tested here and should not be read as confirmed by this page: the count of vendors taking a margin on model tokens, and the 52-page unit-invention count. Both were measured across a wider frame on a different date. What this census does confirm within its own 21-vendor frame is that a named billing unit with no rate attached is still the most common thing a pricing page carries, at 6 of 16 pages.

6. Limitations

• •Prices change and this is a snapshot. Every row states what was on the URL on 10 September 2026. A vendor that publishes nothing today may publish tomorrow, and the reverse. The census is refreshed quarterly and each edition keeps its own date.
• •Pages can be geo-varied, A/B tested or rendered client-side. A pricing page can show different tiers by region or by cohort, and a figure loaded by script after the page renders may not be captured by a fetch. The Databricks and Snowflake rows are exactly this case and the note on each says so. A row records what one fetch saw, not what every visitor sees.
• •Contact sales is not evidence of a high price. Nothing here supports an inference about what any vendor charges. A vendor that quotes on request may be cheaper or dearer than one that publishes. This census measures disclosure, not cost, and no row should be read as a judgement about a vendor's price.
• •Websites only, so a marketplace listing will not appear here. Several vendors that publish nothing on their own site publish a contract price on AWS Marketplace. Those listings are deliberately out of frame here and are covered in the August report linked in section 5.
• •Free tiers are recorded but do not make a rate card. A vendor whose only number is $0 is not counted as publishing a rate. Datus AI and Nao Labs both publish free tiers and are classed as contact sales only, because neither shows a price for anything you pay for.

7. Common questions

8. Take the data

The full census, with every row, URL, fetch date and note, is published as JSON under CC BY 4.0 at published-pricing-agentic-data-engineering-2026.json. Cite as: Data Workers, Published pricing in agentic data engineering, 10 September 2026. The vendor frame comes from The Agentic Data Engineering Index, whose dataset is published on the same licence.

• •Clone the open-source core, Apache-2.0, 11 agents and 160+ MCP tools, read, analyse and recommend only: github.com/DataWorkersProject/dataworkers-claw-community
• •Book a 45-minute working session: dataworkers.io/book-demo