Models and spend

The right model, at a cost you control.

A broad choice of models, an allowlist your admins set, and limits by team, member, and client project. Usage is attributed as it happens, so you can cap it or rebill it.

Request a demoSee how we handle your data
Models your firm allowsOrganisation
  • OpenAIAllowed
  • AnthropicAllowed
  • GoogleAllowed
  • MistralAllowed
  • xAIAllowed
  • DeepSeekOff
Auto picks from this list only. Teams can be narrowed further.
Control

Choose the models. Set the ceiling.

Model choice and spend are the same decision. Hebno keeps both in one place, with the numbers your finance team needs.

A catalogue you decide on

Admins enable models for the organisation, then narrow them per team when a client or a jurisdiction requires it. Nothing outside that list can answer.

  • Six providers, so you are not locked to one vendor
  • Team rules act as a whitelist over the organisation catalogue
  • Auto picks the best model from what that conversation is allowed to use
Models your firm allowsOrganisation
  • OpenAIAllowed
  • AnthropicAllowed
  • GoogleAllowed
  • MistralAllowed
  • xAIAllowed
  • DeepSeekOff
Auto picks from this list only. Teams can be narrowed further.

Hard and soft limits, at four scopes

Set a ceiling on the organisation, a team, a member, or a client project. Hard limits block new requests at the cap. Soft limits alert and let the work continue.

  • Alerts fire at the threshold you set, not only at the cap
  • Child limits cannot add up to more than their parent
  • Reserve shows the headroom left before you assign more
Budgets and limits
  • Organisation · monthly
    $18,400 / $25,000
    74% usedHard limit
  • Creative · monthly
    $4,200 / $5,000
    84% usedHard limit
  • Claude · monthly
    $2,100 / $3,000
    70% usedSoft limit

Cost attributed to the client project

Every turn is recorded against the model that answered, the member who asked, their team, and the client project it was tagged to. Rebilling stops being an estimate.

  • Usage visible by team, member, and client project
  • Underlying provider cost available to finance and admins
  • Monthly periods that reports and limit checks share
Cost by client project
ClientProjectCost
NorthwindBrand refresh$1,840
Atlas LegalContract review$3,210
MeridianQ3 research$920
HalcyonSupport pilot$470
How it works

From catalogue to invoice line.

  1. Step 1

    Enable the models

    Admins turn models on for the organisation, then narrow them per team where an engagement calls for it.

  2. Step 2

    Let Auto choose

    Auto picks a strong model for hard problems and a lighter one where that is enough, only from what is allowed.

  3. Step 3

    Set the limits

    Cap spend at organisation, team, member, or project level. Hard limits block, soft limits alert.

  4. Step 4

    Attribute the spend

    Usage is recorded as it happens, so a client project's AI cost is a number, not a guess.

Commercials

How Hebno charges.

Pricing is a proposal scoped to your team, your workflows, and the sources you need. The commercial model underneath it is simple.

Seats plus included credits

A flat platform fee per assigned seat. Each seat includes a monthly credit allowance, and credits measure AI work.

No markup after the allowance

Usage beyond the included allowance is billed at provider cost with no Hebno markup. Bring your own keys and there is no Hebno usage fee.

Provider cost for finance

Admins can see the underlying provider cost for rebilling, reconciliation, and margin work on your side.

Monthly periods, in UTC

Limits reset on UTC calendar months, and reports use the same boundary as the limit checks.

Members see credits

Day to day, people see credits rather than token counts or provider rate cards.

Your rebilling policy stays yours

What you charge a client on top of the underlying cost is your decision, not our take rate.

The rest of the platform

One loop, not four tools.

Research, drafting, sources, and cost sit in the same workspace, so work does not have to be moved between them.

Deep research

Every claim carries its source.

A visible work trace, sources labelled internal, client, or public, and each finding marked supported, inferred, or uncertain before anyone reviews it.

Explore
Editor

Finish the draft without losing the evidence.

Edit a deliverable in place with citations attached, rewrite a selection, check the source behind a line, then export to Word or PDF.

Explore
Integrations

Works from the systems you already use.

Connect Google Drive and OneDrive, pull files into a chat or a project, and keep working inside Word and PowerPoint.

Explore

See what a month of AI actually costs your firm.

Book 20 minutes and watch a team go from brief to source-backed draft, with every source visible and ready to review.

Request a demo