# Pricing

Free tools and pay-as-you-go provider tools, with a prepaid workspace balance.
## Discovery and zero-price tools

Discovery and inspection do not start a provider execution. A zero-price tool can be executed only with an authorised workspace, within the applicable allowance and limits.

## Prices come from the current tool contract

Each tool publishes its pricing model, unit and relevant reservation limits. Paid tool prices have no Agentik markup. Inspect the actual tool before running; a sample input is not a quote for every possible request.

## Adding credit

The backend publishes a 5.5% platform fee on a credit purchase, with a minimum fee of $0.80 and a minimum credit purchase of $5.00. The fee is added to the requested credit. Currency: USD. Payment mode in this environment: **live**; payment availability: **enabled**. Automatic recharge: **unavailable**.

A human completes a credit purchase through the payment flow when it is enabled. The agent does not add money by executing a tool. Checkout shows the applicable total before confirmation.

## Reservation and settlement

Before a paid run, Agentik reserves the quoted maximum. The final charge is based on the tool’s measured consumption, up to that run’s reservation; unused reservations are released. Partial consumption can be billable even if the final result is incomplete or the connection fails. An uncertain outcome can retain its reservation pending reconciliation.

## Read the receipt

Use the returned run ID to retrieve its state and charge. Do not interpret an error as an automatic refund. Keep the same idempotency key when retrying the same intention, and check an existing run before starting another. [Error recovery](/docs/errors) · [Idempotency](/docs/idempotency).

## Questions

### Is there a spend cap per agent or per key?

Keys can carry hourly, daily, lifetime and per-tool spending caps. Most tools reserve their maximum from your prepaid balance. Tools explicitly marked actual-cost reserve an estimate and bill the exact provider receipt, which may exceed the reservation or balance; capped keys cannot run those tools. A negative paid balance must be funded before new calls.

### Are prices really the providers’ prices?

Tool calls have no markup. Each tool publishes its pricing source and billing unit. For providers sold in credit bundles, the listed rate converts credits to USD using the reference plan shown in the tool pricing. Your agent can inspect the current rate and estimate the cost before execution. Your first successful top-up has no platform fee when you add $10 or less in credits. The standard fee applies to larger first top-ups and all later top-ups.

### How do you handle rate limits across the underlying APIs?

Agentik applies provider and API-key rate limits. A rate-limit response identifies the limit and how long to wait. Calls rejected before dispatch do not consume provider credits. For an uncertain upstream outcome, Agentik keeps the reservation until consumption can be confirmed.

### Can developers set spending limits per agent or per tool?

Yes: a daily cap per key, a list of allowed tools per key (exact tool or whole provider), and a daily cap per tool on that key.

### How do you decide which tool is the best match when several are available?

Discovery returns relevant tools with their descriptions, price estimates and available performance information. Your agent can request a supported cost, speed or reliability preference, then inspect the selected tool before executing.

### Do you support x402?

No, by design. There is no crypto and no wallet: money comes in through a human on a top-up page, the agent spends a balance. If an agent runs out of credit, the response carries a link only a human can complete; MCP clients that support URL elicitation open it directly.

[Current pricing JSON](/pricing.json) · [Tool catalog](/tools.md)

---

[HTML page](/pricing) · [Agent index](/llms.txt)
