Use cases · Directors of delivery and account managers
Quote a new project from what similar work cost.
Every closed month records what each ticket really cost. When a client asks for new work, you start from those numbers.
What you need before you send a quote.
- What will the agents cost on this new project?
- Which past work is it like?
- Did the last quote hold?
From past tickets to a token budget.
A token forecast compares the new ticket's title and description with the summaries of past work, shows the matches with a confidence, and proposes their median cost.
- 1
Close a few months
Each closed period adds what tickets really cost. The more periods, the better the forecasts.
- 2
Add the new tickets
Sync them from Jira Cloud, Azure DevOps or GitHub Issues, or create them by hand.
- 3
Read the token forecasts
Each new ticket gets a token forecast from similar past tickets. It is a suggestion.
- 4
Accept and quote
An accepted token forecast becomes a token budget. The quote to your client is yours to write.
A quote you can defend.
Your own baseline
Forecasts come from your organization's own past tickets, never from other customers.
Agents do not guess
Agents never estimate their own cost. Their self-estimates match actual cost poorly, with a correlation of about 0.4.
Check it afterwards
Once the work is done, the quote's token budget sits next to actual on every ticket.
In the docs
Other questions Token Controller answers.
- Agency owners and account managersPass AI agent cost through to your clients
- Controllers and financeClose the month against your provider bills
- Project managersToken budgets per project and ticket
- Engineering managers and heads of departmentA token budget for each person
- Teams with support, security and other agentsAgents that don't write code
- Teams that keep data in-houseSelf-host Token Controller
Start alone in a few minutes. Add your team when the first month closes.
No card for the free plan. Fourteen days to try the Team plan.