Session sizing calculator
Size your deployment in 60 seconds
Wren AI self-hosted comes with unlimited users and agents. The license is sized by concurrent sessions, the questions in flight at your busiest moment, whether your team asks the GenBI agent or Claude calls Wren over MCP. Tell us your team and we'll estimate the sessions you need.
- Unlimited
- users on every deployment
- 0
- per-seat fees
- ≈ 5
- active users per session
- 1
- pool for UI, API, Slack, Teams & MCP
Your users, by usage
40 users totalOccasional use, Ad-hoc questions and the odd dashboard, a few times a week.
Daily self-serve, Self-serve analytics through the working day. Our default guide.
Always-on, Constant querying, embedded GenBI apps, Slack/Teams bots and AI agents calling over MCP or the API.
Recommended capacity
sessions
Sized for 40 users, with 20% headroom.
The self-hosted Wren AI platform, licensed by concurrent sessions and deployed in your own cloud or infrastructure.
Enterprise Plus is licensed by concurrent sessions, sized to your team. Ask sales for a quote and we'll confirm the exact price for your 9 sessions on a quick call.
Estimate for guidance only. Final pricing confirmed by Wren AI sales.
What actually scales your license
Everyone else on the org chart, and anyone you add later, comes free. Only peak load moves the number.
- A per-seat tool would bill
- —named users
- Wren AI bills
- —concurrent sessions
How concurrent sessions compare
Most analytics platforms bill by named user or by metered usage. Blue checks mark the model that works in your favour.
- Works for you
- Depends
- Works against you
- What you pay for
- Every named user
- Add 100 occasional viewers
- +100 licenses
- API, Slack, Teams, MCP, embedded
- Extra SKUs or add-ons
- Budget predictability
- Grows with headcount
- Maps to your infrastructure
- No
- What it rewards
- Restricting access
- What you pay for
- Every question
- Add 100 occasional viewers
- Bill rises with usage
- API, Slack, Teams, MCP, embedded
- Every call metered
- Budget predictability
- Varies month to month
- Maps to your infrastructure
- Loosely
- What it rewards
- Asking less
- What you pay for
- Peak load only
- Add 100 occasional viewers
- +0 to a few sessions
- API, Slack, Teams, MCP, embedded
- Included
- Budget predictability
- One annual line item
- Maps to your infrastructure
- Yes, sessions size GPUs
- What it rewards
- Giving everyone access
Wren AI Cloud is credit-based, which suits teams starting small without infrastructure. Self-hosted trades that for a fixed annual license and unlimited users. Compare Cloud and self-hosted
Sessions track compute. Seats track org charts.
A concurrent session is one LLM-driven request in flight. While it runs it occupies inference capacity, so the number of simultaneous sessions is the same number that sizes your GPUs or your model endpoint. A named-user license charges for accounts, most of which are idle at any given moment. A concurrent-session license charges for the resource your deployment actually consumes. That's why the price scales with load, not with how many people you invite.
Anyone in the company, plus partners, embedded end-users and agents.
Questions being answered at your busiest moment.
GPUs or model endpoint, sized for the same peak.
Anyone in the company, plus partners, embedded end-users and agents.
Questions being answered at your busiest moment.
GPUs or model endpoint, sized for the same peak.
What counts as a concurrent session?
A concurrent session is a single LLM-driven request to Wren AI, answering a question or producing an analytical output (chart, summary, report or GenBI App), counted from the start of the request to the return of its output. It's the same whether the request comes from the UI, the API, Teams, Slack or an MCP endpoint, so there are no separate per-call API charges. AI-assisted authoring tools like the Modeling AI Assistant and AI Advisor don't count. As a rough guide, one concurrent session supports around five active users.
1Ask from anywhere
- UI
- API
- Slack & Teams
- AI agents (MCP)
2One session pool
3 of 5 in use
Your license = the number of slots
start to finish = 1 session
3Get an output
- Answer
- Chart
- Report
- GenBI App
Questions from evaluations
Concurrent-session licensing, answered
Because sessions measure what the system does, and seats measure who has an account. Each concurrent session is an LLM-driven request occupying inference capacity, so the session count is the same number you use to size GPUs or a model endpoint. Named-user pricing charges for accounts that sit idle most of the time and pushes teams to restrict access to control cost. Concurrent sessions let you give everyone access and pay for peak load.
Nothing, on the license. Users are unlimited. What matters is how many questions are in flight at your busiest moment. Adding a few hundred occasional viewers usually moves peak concurrency by a handful of sessions, and often by none, because their questions spread across the day.
No. Every channel draws from the same session pool, so there is no separate per-call API charge, no embedded-analytics SKU and no per-user add-on for chat integrations. A question asked through an AI agent on the API counts exactly like a question asked in the UI.
You add sessions to the license, and sales quotes the increment. There's never a true-up on user count, and large-volume purchases are discounted. Many teams start from the estimate on this page, confirm real peak concurrency during a pilot, and license that.
Ready to size it for real?
Take your estimate into a quick call. We'll confirm the exact price for your sessions and the infrastructure it needs.