Know what every agent costs to run

"What does the bot cost us?" is the question that decides whether an agent program grows or gets quietly shut down — and most platforms can only answer it with a provider invoice four weeks later.

Botlit answers it per run. Every execution records its token usage and computed cost alongside the model that ran, so cost attribution is native: by agent, by app, by workspace, by time window. The unit economics of "support deflection bot, last week" is a filter, not a finance project.

For agencies, this is pricing infrastructure. With one workspace per client, per-client cost is structural — Marco's team can see which clients' bots are heavily used, what each costs to serve, and where margin is leaking, then decide deliberately: move the high-volume FAQ traffic to a cheaper model binding, keep the complex escalation agent on the premium one. Each change is an agent version with a changelog, so the cost optimisation itself is auditable.

For internal platform teams, the same records defend the budget. When finance asks why the LLM line item doubled, the answer is specific: this agent, this usage growth, this cost per resolved conversation — with the execution ledger to back it.

Botlit rides the Burdenoff platform's billing for plan management and subscriptions. Automated quota enforcement — per-plan caps that gate usage at the metering layer — is on the roadmap and labelled as such; the per-execution cost ledger you would want underneath any quota system is already writing, on every run, today.