Seat Plus Usage Pricing: Examples & Companies

What is it

Seat Plus Usage Pricing is a subset of hybrid pricing where a per-user seat fee is combined with usage-based charges that typically dominate the bill at scale.

The distinguishing feature is which component grows. In a generic hybrid model, the fixed fee could be a platform fee, a base subscription, or anything recurring. In seat-plus-usage, the fixed fee is explicitly priced per user — you pay for every engineer, agent, or support rep with access — and the metered component (tokens, credits, resolutions, compute) is the part that scales with consumption. For heavy customers, the usage line usually exceeds the sum of seat fees.

The pattern is the default in AI developer tools. Cursor sells a $20/month Pro seat that ships with $20 of API usage, GitHub Copilot pairs a $10–$39 per-seat subscription with a GitHub AI Credits pool billed at $0.01/credit, and Augment Code prices per-developer seats ($60 Standard, $200 Max) that each bring a fixed credit allotment. Outside coding, Intercom Fin meters resolved support tickets on top of a seat plan, and Vercel layers a multi-dimensional usage ledger on top of a $20/seat Pro fee.

It solves a real problem: AI inference is a variable cost that rises with how intensively each user works, so a flat per-seat fee alone either loses money on power users or overcharges light ones. Bundling a usage pool into the seat lets vendors land teams on a predictable line item while still recovering the cost of heavy consumption through metered overage. The risk — covered below — is that the same mechanic produces bill shock when adoption ramps faster than the buyer modeled.

How it works

Every seat-plus-usage bill has two layers: a recurring per-seat charge that grants access, and a metered layer that counts consumption. The design decisions are all about how those two layers interact — what the seat includes, how usage is denominated, and what happens when the bundled allowance runs out.

The denomination choice is the most consequential design decision, and the corpus splits cleanly. Cursor denominates its pool in dollars, so the seat fee is effectively a pass-through wrapper around raw API cost. GitHub Copilot abstracts to a credit at a fixed $0.01 conversion, and GitLab uses Duo Credits priced at $1 each. Intercom Fin skips compute units entirely and meters business outcomes at $0.99 per resolution. Each abstraction trades transparency for predictability differently.

Unit math: Total bill = (seats × seat_price) + max(0, usage − bundled_allowance) × unit_price. For a 10-seat Cursor Teams account ($40/seat) that burns $3,500 of model usage in a month against its bundled $20/seat pools: (10 × $40) + ($3,500 − 10 × $20) = $400 + $3,300 = $3,700 — the usage line is 8× the seat line, the “usage dominates at scale” signature.

Companies using this

Thirty companies in the corpus combine a per-seat fee with metered usage, spanning AI coding tools (Cursor, GitHub Copilot, Augment Code, Bito, Sweep AI, Claude Code, Replit AI), DevSecOps and infra (GitLab, Vercel, V0, Mintlify), enterprise productivity (Glean, Notion AI, Writer, Dust), automation (Automation Anywhere, UiPath AI, Cognition), observability (LangSmith, Weights & Biases), customer support (Intercom Fin, Zendesk AI), sales (Apollo.io), and creative and content tools (Leonardo.ai, Typeface, Copy.ai). The table below shows each company’s seat and usage structure.

Patterns observed

Across 30 companies the seat-plus-usage structure is the dominant packaging choice in AI product categories where marginal cost is real and variable. The AI coding cluster is the most concentrated example: Cursor, GitHub Copilot, Augment Code, Bito, and Sweep AI all pair per-developer seats with credit or dollar pools because inference cost per developer spans 5–10× depending on model selection and agentic depth. Cursor’s own benchmarks capture that spread: daily Tab users stay within their $20 pool, daily Agent users typically run $60–$100/mo, and power users running multiple agents often pass $200/mo on the same seat. The same logic applies at the enterprise productivity layer — Notion AI and Writer sell per-seat access and meter AI credits on top because LLM inference is a variable COGS a flat seat fee cannot absorb.

The denomination of the metered layer reveals the vendor’s philosophy — from Cursor’s raw-dollar pass-through to Copilot’s abstract credit to Fin’s per-outcome charge (all detailed above). Two categories the mechanics table doesn’t cover extend the same idea: Automation Anywhere and UiPath AI measure AI consumption in bot licenses and automation credits anchored to RPA workflows, choosing a unit their RPA buyers already budget against. Each abstraction is designed for a different buyer persona and a different cost structure.

The seat-plus-usage pattern has also colonized categories where it wasn’t historically expected. LangSmith and Weights & Biases both sell per-seat platform access and meter trace or evaluation volume on top — W&B’s Pro tier starts at $60/mo and adds meters on storage GB, Weave ingestion, and per-token Inference. Leonardo.ai and Typeface sell per-seat subscriptions and meter image or brand-content generations. Mintlify prices per developer seat and meters AI-enhanced documentation usage. The 30-company breadth shows the model has become the default packaging for any AI product where output intensity varies significantly across the user base.

A structural pattern across the corpus: self-serve seat ceilings consistently funnel into sales-gated enterprise. Writer makes the boundary explicit — self-serve Starter caps at 5 seats before the sales-quoted Enterprise platform takes over. The transition from transparent self-serve pricing to opaque enterprise quoting happens at different team sizes but is near-universal, because the usage-dominates-at-scale property creates negotiable leverage both parties want to handle out of public view.

Counterexamples & variants

Intercom Fin is the cleanest variant at the boundary: technically it fits seat-plus-usage (Intercom platform seats at $29 Essential, $85 Advanced, and $132 Expert plus a metered layer), but its meter is outcome-based, not compute-based. The $0.99 per resolution charge is the only usage unit in this set tied to a business result rather than tokens, credits, or compute — and Fin can even run standalone with no seats required. It shows that seat-plus-usage and outcome-based pricing are not mutually exclusive: the seat covers the human support suite, the outcome charge covers the AI. Zendesk AI runs a similar hybrid — per-agent Suite seats at $19–$115/seat plus AI-resolved conversation usage layered on top.

Glean is the place the model breaks down for buyers. It runs the per-user Enterprise Flex seat plus pooled FlexCredits structure, but every figure is quote-only — glean.com/pricing 301-redirects to the homepage. FlexCredits do have documented burn rates (a Thinking Mode query on premium models consumes roughly 35–120 credits, slide generation 45–142), yet with no public seat rate and no public credit price, buyers cannot model the usage-dominates-at-scale risk that defines the category. Poolside and Cognition show the same opacity at the frontier AI coding layer — both are functionally seat-plus-usage above their trial tiers, neither publishes a full enterprise rate card. The structure is seat-plus-usage; the transparency that makes it evaluable is absent.

Vercel is the high-complexity variant: a per-seat Pro fee at $20/seat sits on top of a multi-dimensional usage ledger (Fast Data Transfer, Edge Requests, function invocations, Fluid Compute Active CPU, and more), with viewer seats free and the first developer seat included. V0 runs a related structure within the Vercel workspace. Together they demonstrate that “usage” in seat-plus-usage need not be a single unit — at the infrastructure layer it can be a multi-dimensional ledger, which makes the bill far harder to forecast than a single dollar or credit pool. Teams building on LangSmith face a similar multi-meter challenge: seat access plus per-trace billing across two retention classes on the same invoice.

The most interesting inversion is Copy.ai, which runs a flat per-seat subscription for its entry Chat plan (seat-only) but layers workflow-credit and per-word metering onto its GTM tiers, with Enterprise sales-quoted. It shows that “seat-plus-usage” is not a fixed direction of travel; in content tools where volume variance is enormous, vendors experiment with which tier is seat-only and which unlocks the variable meter. Claude Code pushes the boundary the other way — it offers two paths, a seat bundle inside Claude Pro/Max/Team subscriptions with usage limits, or pure pay-as-you-go Anthropic API billed per token — sitting right on the seam between seat-plus-usage and pure-usage.

What this means for buyers vs vendors

For buyers

Model expected usage against the bundled allowance before you sign — the seat fee is the floor, not the bill, and at real team usage the overage line runs several times the seat line. A Notion AI Business team generating AI content heavily can exhaust its bundled credit pool fast. Ask for the overage unit price, whether unused allowance rolls over, and whether spend caps or alerts exist. For dollar-denominated pools like Cursor’s you can forecast precisely; for abstract-credit pools like GitHub Copilot’s ($0.01/credit) or Augment Code’s (per-model burn rates), get the credit-to-task conversion in writing; for quote-only structures like Glean’s or Cognition’s, insist on a worked example at your projected seat count.

Enterprise buyers negotiating annual contracts with UiPath AI, Automation Anywhere, or Writer face a harder version of the same problem: the base seat is a committed cost, but the AI consumption component is variable across a workforce of hundreds. Negotiate usage caps or monthly spending alerts at contract time. The usage invoicing & billing cycles guide covers how overage timing affects cash flow, and the thresholding & alerting guide explains how to govern variable spend before it surprises you on invoice day.

The credit denomination also changes how you audit the bill. When GitLab prices a Duo Credit at $1 or Copilot at $0.01, that fixed conversion is a commitment you can reconcile against usage logs — insist on line-item reporting per seat. Where the meter is outcome-based, as with Intercom Fin’s $0.99 resolution, confirm exactly what counts as a billable event (a resolved conversation vs. any Fin interaction), because the definition drives the whole invoice. The prepaid credits models guide walks through how bundled pools, rollover, and top-ups interact so you can pressure-test a vendor’s credit design.

For vendors

Seat-plus-usage fits when your marginal cost per user is real and variable — AI inference, compute, or per-outcome delivery — and when buyers expect a predictable per-seat line to anchor procurement. Decide your denomination early: dollar pass-through like Cursor’s maximizes trust but exposes you to model-cost swings, while an abstract credit like GitHub Copilot’s or GitLab’s insulates margin at the cost of transparency. The corpus’s conversion arc is instructive: GitHub Copilot, Augment Code, and GitLab all migrated from flat or “unlimited” seat add-ons to credit metering once the model bill arrived — GitLab moved Duo from flat $19/$39 per-seat add-ons to $1-per-credit metering — and each migration carried trust cost. Build credit metering into the product from day one rather than bolting it on after users experience the pricing change.

You will need metering infrastructure that can aggregate usage per seat, enforce the bundled allowance, pool credits at the team tier, and bill overage cleanly — start with the introduction to usage-based pricing and model scenarios at multiple usage intensities with our pricing calculator. The 30-company corpus also shows the organizational boundary: seat-plus-usage scales through self-serve for individual developers and small teams (see Writer’s 5-seat Starter cap), then requires enterprise sales for large accounts where the usage component becomes a meaningful negotiation. Plan the self-serve ceiling deliberately.

Explore this theme in the knowledge graph →