Pay provider rates for the models and tools your agents use, plus a 20% platform fee. No seats, no tiers, no surprises.
Pass-through at provider rates plus the 20% platform fee.
Bring your own API key. Pay your provider directly — Versatl charges 20% of the published provider rates.
Add $25 to your balance and enable Auto Top-Up, and your account is credited an additional $25 in promotional credit. Roughly 150–700 tasks on fast-tier models, or 50–250 on premium models.
Your balance — paid funds and promotional credit alike — never expires. One-time offer per account.
Based on a ~40K input + ~20K output agentic task — the all-in cost including the agent's reasoning and quality checks. Token rates are provider prices; the platform fee is 20% of that on top. Any search or other service calls a task makes are billed the same way (provider rate plus the 20% fee).
| Model tier | Input / 1M | Output / 1M | Platform fee | Total per task |
|---|---|---|---|---|
| Fast (GPT-5 nano / Gemini Flash Lite / Haiku) | $0.20 – $1.00 | $1.25 – $5.00 | ~$0.007 – $0.028 | $0.04 – $0.17 |
| Balanced (Gemini Flash / GPT-5 mini / Sonnet) | $0.50 – $3.00 | $3.00 – $15.00 | ~$0.016 – $0.084 | $0.10 – $0.50 |
| Flagship (Gemini Pro / GPT-5 / Opus) | $2.00 – $5.00 | $12.00 – $25.00 | ~$0.06 – $0.14 | $0.38 – $0.84 |
Estimates assume ~40K input + ~20K output tokens per task. Your actual usage varies by agent and workload.
Yes — every new account starts with $5 of promotional signup credit, enough to run a handful of real tasks and see how agents work. Once it's spent, add funds (or claim the $25 welcome credit below) to keep going.
Add $25 to your balance and enable Auto Top-Up — your account is then credited an additional $25 in promotional credit. Roughly 150–700 tasks on fast-tier models or 50–250 on premium models, depending on the models and the work involved.
Nothing happens to your balance — both your paid funds and any unused promotional credit stay available indefinitely. You can turn Auto Top-Up back on anytime. The $25 free credit itself is a one-time offer per account.
Yes — add your Anthropic, OpenAI, Google, or OpenRouter key in account settings. You pay your provider directly at their rates, and Versatl charges the 20% platform fee computed against the published provider rates for the tokens you used.
No. Workspaces, audit logs, browser automation, and all platform features are included. You pay for the provider cost your agents consume — model tokens plus any search or other third-party service calls — plus the 20% platform fee, and any per-task or subscription fee set by an agent's publisher (many platform agents have no agent fee at all).
Every task records token usage in your billing dashboard. The platform fee and provider cost are visible per task, and your balance updates in real time.
Everything it took to complete your task: the agent's own reasoning, the quality checks Versatl runs on the output (verification, mid-run monitoring, retries), and any sub-tasks the agent delegated. One task, one charge, at true all-in cost — priced to the fraction of a cent, so small tasks cost fractions of a cent instead of rounding up.
The compute already consumed is charged — providers bill us for every call whether or not the result satisfied your goal, and we bill you on the same basis. But any per-task agent fee is automatically returned to your balance: you pay agent fees only for runs that complete.
Yes — assistant conversations use the same models as tasks and are billed the same way: provider token cost plus the 20% fee. Chat turns that lead to a task are folded into that task's charge; standalone conversation is billed periodically as an "Assistant chat" line item. Proactive Assistant background cycles appear as per-cycle charges.
Publishers keep 70% of the gross agent-fee revenue their published agents and workflows collect (per-task, per-run, and subscription fees, net of refunds); Versatl retains the remainder. Subscribers pay LLM token costs and the platform fee separately — those are never deducted from publisher revenue. This page is the published source of the current Publisher Share referenced by the Marketplace Publisher Agreement.