Pricing and metering

TalkIDE has two independent billing streams: AI credits (prepaid) and hosting (postpaid). They are metered, capped, and invoiced separately.

AI credits

AI credits pay for agent work: every conversation turn, code generation, review, and deployment orchestration that the AI team performs.

How it works

Credits are a prepaid balance. You top up in advance; the platform deducts usage as agents work. If the balance runs out, agent work stops until you top up again.

Topping up

Top up from the Billing section of your profile. Minimum $5, maximum $500 per top-up. Payment is processed by Stripe; credits are applied once the payment is confirmed.

Spending limit

You can set a spending limit on your account. When the balance reaches the limit, agent work stops before the balance is fully exhausted. The limit is a soft cap. Mara will notify you before work is paused. The default is no limit (agents work until the balance is zero).

Hosting

Hosting charges cover the compute, database, storage, and network resources your published apps consume on TalkIDE infrastructure.

How it works

Hosting is postpaid and usage-based. There are no fixed tiers or quotas. You pay for what you use. Usage is measured hourly via OpenCost, a Kubernetes-native cost metering tool. A TalkIDE markup is applied to the raw infrastructure cost.

Trial period

Every new published app starts with a 14-day trial from the date of first publish. No hosting charges accrue during the trial period.

Invoicing

Hosting costs accrue throughout the month and are invoiced on the first day of each calendar month via Stripe. If a Stripe charge fails, TalkIDE retries up to three times. After three failures, the account enters a 7-day grace period. If payment is not resolved within the grace period, the account is suspended and published apps are scaled to zero.

Spending cap

You can set a hosting spending cap on your account. If accrued charges reach the cap before the invoice date, new usage accrual stops for the rest of the billing period. The cap resets at the start of the next period.

Measured resources

Hosting cost is derived from Kubernetes resource allocation data (CPU, memory, storage) as reported by OpenCost. Measurements are collected hourly. The per-hour cost for each running app is calculated from its allocated pod resources and the TalkIDE hosting rate.

Apps that have no active deployment (including archived projects) do not consume hosting resources and are not charged. A preview deployment counts the same as a production deployment: both run as pods in your namespace and accrue costs.