Pricing & limits — what's free and what's metered
Before any price table, the thing to know is what gets metered. London System Agent is in beta: building and validating an agent is free, and cost only starts when you run live models in Real mode.
What you can do for free
With no sign-up, you can open the studio, compile a one-sentence requirement into a graph, run a Stub-mode simulation, check structural validation, and export a runnable bundle — all with no token cost. Stub mode calls no live LLM, so you can iterate on structure, branching, and loop termination for free.
Sign in with a Google account to save your agents to the library, and a small one-time signup credit lets you try a Real run without paying.
Where cost starts
Cost only starts in Real mode. Real mode calls live LLM APIs through your own provider keys (BYOK), and the provider bills your account directly. The platform only runs within prepaid credits and a budget cap.
Per-run cost is shown transparently in the cost panel with per-step input/output tokens, and the same cost and status are overlaid on the graph nodes — so you can see exactly which node spent what.
How to keep cost under control
The default strategy is simple: iterate for free in Stub while shaping structure, and switch to Real only to check quality. Set a budget cap and the runtime ledger accumulates cost step by step, stopping any run that exceeds it.
An honesty note: the cost panel's pricing table is based on approximate public list prices — it is not a ledger that reconciles 1:1 with a provider invoice. Treat it as an optimization aid and budget alarm.
If you need more
During beta you start with a free agent per user. If you need more — more capacity, a dedicated limit, or team usage — reach out at totaro@totaro.co.kr. Formal plans will be published as usage grows.
Your turn to build
Compile your first agent graph from a single sentence.