The devtools company with one product is getting rare. Between mid-August and early November 2026 a database added a gateway, a function runtime, auth and live queries; an editor added a git forge; a VPN added an AI governance product; a cloud added an agent runtime and priced it twice; a frontend cloud added CI. What this thread follows is what the second product does to the pricing page. Read on 2026-11-07, the answer is one shape with two halves: the new product’s access rides the plan the customer already has, and its usage goes on a meter of its own. The meter does not wait for the beta to end, and the incumbents that look like they bundle mostly bundle usage too. A pack price set against the parts survives at one older incumbent, JetBrains, and one acquirer, xAI.

The first beta to bill billed on its own meter

Vercel CI, which runs only the tasks a change affects and reports as a GitHub check, launched on 2026-10-28 “in Beta today on all plans”. The launch post said nothing about price, but Vercel’s CI docs, dated the same day, did: it bills CPU minutes, with 200 a month included on the free Hobby plan, $0.0035 a CPU minute on Pro and custom pricing on Enterprise. The same row, marked “Beta”, sits in the Build & Deploy section of Vercel’s pricing page. It was real money from the start: on 2026-11-02 a Pro user reported being billed 20 to 30 minutes for tasks that ran 10 to 20 seconds, and Vercel staff replied the next day that a bug had been fixed and credits applied to affected customers. An earlier version of this thread listed Vercel CI among the second products with no price; that was wrong.

So the first second product in this set to start billing did it the cohort’s way: a small allowance on the free plan, and every paid minute on a meter of its own beside the builds meter it mirrors. Vercel’s one fixed-price add-on moved the same direction. Token Pass is still $30 per team per month on top of Pro for $80 of credit on three open-weight models, but since 2026-11-06 its credit applies to standard API keys, which “fall back to your team’s AI Gateway credits” when it runs out. The subscription is now a prepaid layer over the meter.

GitHub put Copilot in the seat and the usage on a meter

From 2026-10-28, GitHub Enterprise Cloud customers who work with a GitHub or Microsoft account team can give every licensed user Copilot, with the same features and models as Copilot Enterprise and “no Copilot seat charge”. What the seat does not include is any usage: there are “no monthly AI credits per seat”, and usage is metered from the first request against the enterprise budget. Code completions and next-edit suggestions keep working with no budget; chat, code review and Copilot in the cloud stop when it runs out, and the seat stays. Users without an Enterprise licence can buy the same access as Copilot Standalone at $10 a user a month, also with no included usage. Self-serve customers are not eligible; on 2026-11-07, ten days after GitHub promised dates “in the coming weeks”, its docs still said “Currently unavailable”. The Enterprise seat is still $21 a user a month.

It looks like a bundle and is not priced like one. Copilot Business ($19, 1,900 pooled AI credits a user a month) and Copilot Enterprise ($39, 3,900) remain available, and at GitHub’s $0.01 a credit those seat prices equal their allowances. Folding Copilot into Enterprise removes a seat fee that was prepaid usage: a user who would not have used the allowance pays less, and the same usage costs the same. Nothing is discounted for buying the two together.

Atlassian’s collection adds usage, not a stated discount

Atlassian is the incumbent this thread used to file beside JetBrains as a company that already sells a bundle below its parts. Its own pages do not say that. Its billing documentation says a Collection is “one price for the entire collection”, one line item, with each user counted once across the apps and “higher usage limits for meters than in standalone plans”. Its licensing page prices the Teamwork Collection through a calculator and makes no savings claim; what it does claim is that the Collection “provides significantly more Rovo credits than single app subscriptions”. The Rovo documentation gives the numbers: every paid Jira, Confluence and Collection cloud plan includes Rovo, with 25, 70 or 150 credits a user a month on Jira or Confluence (Standard, Premium, Enterprise) against 250, 700 or 1,500 on a Collection, pooled across the organisation, at $0.01 a credit past the allowance, with extra-usage billing from 2026-12-03.

Counting a user once can still make a Collection cheaper than the apps bought apart for a team that uses all of them; Atlassian does not publish the comparison, so this thread does not make it. What Atlassian does publish is the same shape as GitHub’s: AI access in every paid plan, a cent a credit past the allowance, and a bigger allowance as the reason to buy the bigger package.

The same invoice, a meter of its own

This is where most of the 2026 cohort landed. Neon’s pricing page lists every primitive as a row inside its Free, Launch and Scale plans, each with its own rate: object storage at $0.023 per GB-month, functions by capacity-hour and invocation, auth by monthly active users, and the AI Gateway on prepaid credits “no markup”. Since 2026-11-06 the gateway is open to Free accounts too (“You don’t need a paid plan to use it anymore”, card required), and Neon Realtime, launched in public beta on 2026-10-15, now has its own row: “Free during beta” on every plan, with the compute and egress it uses “billed as usual”. Gateway usage, in Neon’s words, “lands on your Neon invoice instead of a stack of lab statements” (2026-08-17). But Neon itself rejects the one-bill label: its GA post says it is building “a set of solid primitives an agent can call”, not “a bundle of managed services behind one bill”.

Cloudflare’s Artifacts, a versioned filesystem pitched at agent-driven development, has its meters published - 10,000 operations and 1 GB-month a month included, then $0.15 per 1,000 operations and $0.50 per GB-month, on the Workers Paid plan only - and says billing begins on 2026-11-12. The same page said 2026-10-14 when this thread read it on 2026-10-21, and Cloudflare’s changelog carries no entry on that move. Moving a published start date is a Cloudflare habit rather than a sign about this product: on 2026-11-02 it moved AI Search’s billing start from November 1 to December 1, saying “Pricing and included usage are unchanged.” PlanetScale’s pricing page gives Neki SKU tables of their own beside Postgres and Vitess (a shard from $30 a month on network storage, Metal from $40), and shows no discount for running more than one of the three.

DigitalOcean says it loudest. Its pricing page reads “Run your agents on one plan, one bill” - and then lists three meters: Harness Runtime from $0.0535 an hour, the Action Gateway from $0.10 per 1,000 calls, inference from $0.05 per million tokens. Agent Droplets (2026-10-01) puts a subscription over them: $5 of trial credit, $50 a month at 15% off agent usage, $200 at 20%, unlimited seats. It is the nearest any 2026 entrant has come to a bundle price, and it stops short of one: the discount applies to DigitalOcean-hosted open models and agent resources only, frontier models stay pay-as-you-go, and Droplets, Spaces and Managed Databases are billed as usual.

No price yet

Three second products still have no price of their own. Cursor’s Origin, a git forge pitched as infrastructure for “the agentic era”, is in early beta, and Cursor’s two pages disagree about who gets it. The Origin page says “all paid plans”; the pricing page, read on 2026-10-21, 2026-10-30 and 2026-11-07, puts “Code hosting via Origin” in the $30 Premium tier and up and in Teams, and not in the paid $8 Plus tier. The product never got a price; it became a reason to buy the tier above. Tailscale’s Aperture, its AI governance product, was announced GA on 2026-08-26, yet its own page still says “We do not have published pricing during the current Alpha and Beta period”; it can be bought without a paid Tailscale plan, and the pricing page names it without an add-on line or a price. Neon Realtime has its row but no post-beta price or date.

A price line of its own

The rest priced the second product apart. Vercel’s Token Pass is the fixed add-on described above. Postman presented four named products as one Agent Stack in August; Postman’s pricing page lists it only under “More from Postman” (as “Astro AI”, linking out) and in no plan table, while astropods.com sells it on its own terms - compute per CU-hour for the size each agent reserves, models at the provider’s price with no markup, free signup credits, Enterprise on request - without mentioning a Postman plan. It is the cleanest shelf in the set: own domain, own meter, own account.

The pack price: one incumbent and an acquirer

JetBrains is the one incumbent here with a published pack price set beside a single product’s: its All Products Pack costs EUR 979 a user a year for companies against EUR 719 for IntelliJ IDEA Ultimate alone, list prices from 2025-10-01 as reported by heise; the pack itself shrank from 11 IDEs to 10 between 2026-05-25 and 2026-06-01 (PricingSaaS’s record of the page).

The one 2026 bundle priced against its parts came from an acquirer. xAI’s X Subscription (2026-10-12) puts Grok, Grok Bot, Cursor and X Premium on one shared usage pool across five tiers, Free to $200 a month, and pitches the $30 Premium tier as SuperGrok, Cursor Pro and X Premium+, “previously $90/month combined” - xAI’s own figure. “Previously” is literal for Cursor: new customers start on the X Subscription, and Cursor staff said on 2026-10-20 that “there’s no Cursor-only self-serve checkout today”. The pool is literal too: Cursor’s help page for Grok Bot, the agent sold with the subscription, says its usage “draws from the same weekly pool as Cursor and the Grok apps, not a separate Grok Bot meter”, so a long agent job spends the same allowance as coding. Existing Cursor subscribers keep their plans “unless you change or cancel”; the switch is optional and permanent, and no end date for the old plans is stated.

xAI was not the first acquirer to put a bought devtool on its own price list. Cognition renamed Windsurf to Devin Desktop on 2026-06-02 (“Your plan, pricing, extensions, and other features remain the same”), and windsurf.com/pricing now redirects to Devin’s pricing page, where one ladder (Free, Pro at $20, Max at $200, Teams at $80 plus $40 a developer seat) sells the cloud agent and, on Teams, “access to Devin Desktop”, with a usage allowance that refreshes daily and weekly. Cognition sets no price against the two bought apart and does not say whether the editor and the cloud agent draw on one allowance. The acquired product became access on the acquirer’s plan, which is the cohort’s shape again.

The adjacent product is a git host

When these companies reached for a second product, three reached for the same one. Netlify built Git into the platform so a project made by its Agent Runners has source control without a GitHub account (2026-08-14). Cursor shipped Origin. Cloudflare put Artifacts into open beta for Workers Paid customers with a contest for developers to build a Git platform for the agentic era on it (2026-10-02). The forge is where an agent’s work lands, so whoever hosts it holds the step after the build. All three entered on an existing plan. Cloudflare’s is the only one with a published meter. Netlify’s pricing page has no priced line for Git hosting, and lists deploys, compute, bandwidth, requests and AI inference in one credit-cost table.

The tension

The positioning moved faster than the price list. “Stack”, “OS”, “runtime”, “backend” - the hero stops naming a product and names a category the company wants to own, and as covered the week of August 17, four of those repositionings landed in one week with no new capability among them. The pricing pages underneath kept selling parts. That is the practical reading for anyone pricing a second product: the 2026 cohort tells the buyer it is one platform and bills it as several meters, which keeps each price legible (the guide’s rule - publish both parts, so the buyer can predict the invoice) and leaves the “platform” claim without a number behind it. Where a single unit does appear, it is a currency or an allowance rather than a discount: Netlify’s credits, DigitalOcean’s agent balance, xAI’s shared pool, GitHub’s AI credits and Atlassian’s Rovo credits each pay for several products at the same rate, and the bigger package buys a bigger allowance. Whether a 2026 entrant prices the bundle, whether Origin, Aperture and Realtime end on meters or inside plans, and whether GitHub’s self-serve version keeps the seat free of usage are the things still open.

What would settle it

4 open loops

  1. 01 By end of 2026

    Does one of the companies that went multi-product in 2026 (Neon, PlanetScale, DigitalOcean, Cloudflare, Vercel, Netlify, Postman, Tailscale) publish a plan holding two of its products at a stated price below buying them apart? JetBrains already sells that way and xAI did on 2026-10-12 with an acquired product; Atlassian's Collections state no saving, DigitalOcean's 15-20% is a discount on agent usage, and GitHub's 2026-10-28 change removed a seat fee rather than discounting a pair. One by year-end makes the bundle part of how the new suites sell; none leaves it an older incumbent's tool.

  2. 02 By as each leaves beta; checked at the end of 2026

    Cursor's Origin (early beta since 2026-08-17), Tailscale's Aperture (no published price "during the current Alpha and Beta period") and Neon Realtime ("Free during beta" on Neon's pricing page, read 2026-11-07) still have no price of their own. Does each leave beta with its own meter or folded into a plan? Vercel CI, the first to bill, did so on its own meter from launch day; Cloudflare says it starts billing Artifacts on its own meters on 2026-11-12, a date it has moved once. Most of the rest on meters says the suite sells the meter; most folded in says it sells the plan.

  3. 03 By when GitHub opens it to self-serve; checked at the end of 2026

    GitHub promised self-serve dates "in the coming weeks" for Copilot included in GitHub Enterprise; on 2026-11-07 its docs still said "Currently unavailable". Does the self-serve version keep the $21 Enterprise seat with no included credits, or does the seat price rise or gain credits? A price rise makes it a bundle priced into the platform; credits make it the old Copilot seat under a new name; neither confirms access-plus-meter as the incumbent's model too.

  4. 04 By when xAI dates the end of the legacy Cursor plans

    Existing Cursor subscribers keep their plans with no end date, and new customers can only buy Cursor through xAI's X Subscription. Does xAI set an end date for the legacy Cursor plans? A date makes Cursor wholly a line in xAI's plan; none leaves two price lists running side by side.

Each is checkable by a date. If one goes quiet, that is itself an answer.