Dust pricing and Dust AI cost per seat on the Pro, Max and Enterprise plans

Dust costs $24 per seat per month for a Pro seat and $120 per seat per month for a Max seat, both on yearly billing. Paying month to month costs 20% more, which is $30 for Pro and $150 for Max. There is also a Free seat with 500 credits that never renew.

All three are seat types inside one plan, which Dust calls Business. An admin assigns a seat type to each person, and you can mix them: most people on Pro, a few heavy builders on Max, an occasional user on Free. Monthly and yearly seats can sit side by side in the same workspace.

Over a year a Pro seat on yearly billing is $288 and a Max seat is $1,440. That five to one gap is the first thing to plan around, because the only thing a Max seat adds is credits. Models, agents and features are identical on both. Dust says so directly: no model is locked behind a higher plan, though more capable models use more credits per message.

Prices are quoted before VAT. US visitors see dollars and European visitors see the same numbers in euros.

A credit is Dust's unit of AI usage. A Pro seat gets 8,000 credits a month, a Max seat 40,000 and a Free seat 500 once. Each message costs token credits, which scale with the model and the length of the work, plus action credits, which are a fixed charge for every tool the agent uses.

Dust publishes the action side as three tiers:

The row that matters for a knowledge use case is the last one. Every time an agent searches your connected data it spends 3 action credits before any token credits are counted, and an agent that checks Slack, then Drive, then Notion to answer one question does that three times. Sub-agents add their own cost to the total.

Two things follow from the arithmetic. First, both paid seats price a credit the same: $24 for 8,000 and $120 for 40,000 are each $0.003 per credit, so Max is not a volume discount, only a bigger bucket. Second, Dust lists programmatic usage on the Business plan at $0.01 per credit, more than three times the rate inside a seat. Anything that runs through the API, a scheduled trigger or an automation tool is billed at that higher rate.

Dust does not publish a table of token credits per model. It shows the cost of each message in the conversation after the fact, so the only reliable way to size seats is to run your real agents for a couple of weeks and read the numbers.

A response that is already being written finishes. After that, a Pro or Max user draws from a shared workspace credit pool if the admin has funded one, and is blocked if not. A Free user is blocked as soon as the 500 credits are gone and cannot use the pool.

Dust's documentation gives the order: individual seat credits first, then the workspace pool, then pay as you go. The third step is Enterprise only. The documentation is blunt about it: pay as you go "is not available on the Business plan", and Business customers "must purchase workspace pool top-up packages in advance".

That is worth reading next to the pricing page, whose FAQ says Pro and Max users can "continue through workspace overage if enabled by an admin up to a capped amount". Both statements are Dust's own. Read together, the overage on a Business workspace is a prepaid pool an admin tops up, never a meter that bills in arrears. Top-ups start at 100 credits, have a maximum per billing cycle and stay valid for one year.

Three settings decide how this plays out in practice, and all three are worth checking before rollout:

• Auto-upgrade seats. When it is on, a member who reaches their limit is moved to the next seat tier automatically, Free to Pro or Pro to Max. A Pro to Max move takes that seat from $24 to $120.
• Default workspace pool limit. With no limit set, every Pro and Max member can draw on the shared pool without a personal cap. Setting it to 0 removes pool access by default.
• The 600 credit checkpoint. Dust pauses a long task started in the web app once it has spent 600 credits and asks whether to continue. It is on by default, the threshold is fixed, and it does not pause tasks started by the API or by sub-agents.

Programmatic usage has no included allowance at all. It draws only from the workspace pool, so an automation that calls Dust has nothing to run on until the pool is funded.

With everyone on a Pro seat billed yearly, Dust costs $2,880 a year for 10 people, $7,200 for 25, $14,400 for 50 and $28,800 for 100. Month to month the same teams pay $3,600, $9,000, $18,000 and $36,000.

Real workspaces are mixed. A 25 person team with 20 people on Pro and 5 builders on Max pays 20 x $288 plus 5 x $1,440, which is $12,960 a year, before any pool top-ups. That is 80% more than the all Pro figure, and it is the more realistic number once a few people start running research tasks and scheduled agents.

The table stops at 100 on purpose. The Business plan is for teams up to 100 people, so seat 101 is an Enterprise conversation with no published price.

The Business plan is capped at 100 people, 3 data connectors, 5 Spaces and 5 remote MCP servers. It includes SOC 2 Type II, US or EU data residency and email support, lists single sign-on as available on demand from 5 seats, and leaves SCIM, audit logs and custom data retention to Enterprise.

The connector cap is the one that changes the budget. Dust connects to more than 20 data sources, but a Business workspace can sync 3 of them. A company whose knowledge lives in Slack, Google Drive, Notion and Confluence is already one over. You can route extra tools through MCP, but each call to an external MCP server is an Advanced action at 3 credits, and there are 5 remote server slots.

Spaces are how Dust segments data and permissions, so 5 Spaces means 5 groupings of who can see what. That is workable for a small company and tight for one with separate finance, HR, legal, engineering, sales and leadership material.

Put plainly, the published $24 seat is the price for a team under 100 people whose knowledge fits in three apps. Everyone else is pricing Enterprise.

Dust does not publish an Enterprise price. The plan is sold by quote on annual terms, and it comes in two shapes: seat-based, with Pro and Max seats as on Business, and pooled, where nobody has an individual allowance and everyone draws on one committed pool of credits.

Enterprise removes the caps that define Business: unlimited connectors and MCP servers, workspace-pooled credits with volume pricing, SCIM, audit logs, custom data retention, single-tenant deployment, a dedicated customer success manager, priority support with an SLA and custom legal terms. It is also the only plan with pay as you go once the pool is empty, behind a spending cap set for your workspace.

Two contract details from the seat documentation are worth knowing before the call. Enterprise seats are annual only, and seats added mid-year are prorated to the end of the term while the new member receives a full credit allowance immediately. And seats vacated when someone leaves are not refunded, they return to your stock for reassignment.

We have not seen a credible published figure for a Dust Enterprise contract, so this page does not give one. Ask for the per credit rate in the pooled plan in writing, because that is the number the rest of the quote depends on.

Dust has a Free seat with 500 credits, but it is a one time allowance and works as a trial. The credits never renew, cannot be topped up, and a member can hold a Free seat only once.

The free option on the Business plan covers Free seats only, for up to 5 members. Offering anyone a Pro or Max seat starts a paid subscription. A person who moves from Free to a paid seat cannot go back, and someone who leaves the workspace and rejoins does not get a second Free seat.

At 3 credits per search of company data, 500 credits is enough to find out whether your agents work on your own content, which is what it is for.

Adding a member mid-cycle is prorated: you pay for the remaining days and the member gets a full credit allowance straight away. Monthly seats can be cancelled at any time. Yearly seats cannot be removed early; they stay in your seat stock until the commitment ends.

When a monthly seat is removed, Dust applies a prorated credit to the next invoice. When a yearly seat is freed, it can be reassigned to someone else at the end of the current credit period, and nothing is refunded. Credits belong to the person, so unassigning a seat does not take back credits already granted for the period, and they are not transferable between users.

The practical advice is the usual one for annual seats with a usage meter. Start people month to month at $30, watch the credit numbers for a cycle or two, then move the stable seats to yearly billing for the 20% saving.

Dust is worth $24 a seat if your team will build and run its own agents: multi-step workflows on a choice of 20+ models, scheduled and triggered runs, shared workspaces, and a builder who owns them. It is a weaker buy if the main need is employees finding answers in company documents, because every search is metered and Business stops at three connected sources.

Where Dust wins is real. It is one of the most flexible agent builders you can buy with a card, it does not lock models behind tiers, it offers EU data residency on the self-serve plan, and it shows the credit cost of every message, which most metered products do not.

Where it costs more than it looks is retrieval. A support team that asks 40 questions a day across Slack, Drive and Notion spends action credits on every lookup, and the fourth app needs an Enterprise quote.

InSearch is built for that second job. An admin connects Google Drive, SharePoint, Gmail, Slack, Notion, Confluence, Jira and Salesforce once, read-only, and everyone gets one answer with a link to each source, filtered by what they are already allowed to see. There is no credit meter and no connector cap: $20, $32 or $49 per user per month billed annually, from three seats, bought by card on the pricing page. Plenty of teams run both, Dust for the agents a few people build and InSearch for the questions everyone asks.