Billing and Credits

Cloud billing, subscriptions, credit packs, and managed-inference usage are commercial surfaces. Fixed prices and included credits are intentionally not copied into these docs because they can change independently of a repository release.

Use the current pricing page  and the checkout UI as the source of truth before purchasing.

Agent chat credits

Agent chat is billed by usage. A chat message can take several model rounds (for example, one to call a tool and one to answer). Each round is billed from the tokens the model actually used, as fractional credits, so a short reply costs less than a long, tool-heavy one. Before a round starts, Genfeed holds an estimate from your balance and settles it to the real amount when the round finishes. To send a message, the balance must cover one round on the chosen model.

Some chat costs no Genfeed credits:

• rounds that run on your organization’s own provider key (BYOK);
• the brand interview; and
• turns that run on your own Claude Code or Codex subscription from the desktop app.

Tools the agent calls, such as image or video generation and publishing, are billed separately at their own credit cost.

Free plan model

Without an active paid subscription, every agent turn in Genfeed Cloud runs on DeepSeek V4 Flash, including sub-agents, plan mode, and recurring agents, whatever model a strategy or setting asks for. Model pickers show a notice with an upgrade link instead of the model list. A round that runs on your own BYOK key for its provider keeps the model you chose. Deployments without organization billing, including Community self-hosts, never lock the model.

How credit amounts are shown

• Balance: whole credits, rounded down, with thousands separators (4,900). From 100,000 credits the balance chip is abbreviated (124k, 1.2M); hover it for the exact value.
• Cost of a message, tool call, or transaction: at most one decimal (0.4, 12.5). Amounts below 0.1 show as <0.1, and zero shows as Free.
• Model pickers: ≈ X credits / message is an estimate for an average message on that model. The actual cost depends on the conversation.

Repository boundary

• The entire repository, billing included, is AGPL-3.0-or-later. Organization billing code lives in the API at apps/server/api/src/collections/{subscriptions,user-subscriptions,subscription-attributions}/ and ships in every image.
• Whether organization billing is live is decided at runtime by hasOrganizationBilling() — see Billing: one build, a runtime gate . Community self-hosts get no-op stubs and no billing routes, and do not sell managed credits locally.
• A Community or desktop client can use managed inference only through an explicit Cloud API key and Cloud-credit authorization.

This page does not promise a trial, fixed price, included quota, seat count, support tier, or SLA.