Billing and credits
How Runbear measures usage, what consumes credits, and what happens when they run out.
Runbear bills on two measures: credits, which agents consume as they work, and interactors, the distinct people who talk to your agents in a month.
Credits#
Every message an agent handles consumes credits. How many depends on two things.
What the agent did. Answering from a knowledge base is cheap. Running a chain of tool calls, reading long documents, or reasoning through a multi-step task costs more.
Which model it runs on. The tiers differ by roughly a factor of three per step, and the difference compounds across a busy channel.
| Workload | Typical range |
|---|---|
| Question answered from knowledge | 10 to 20 credits |
| Task involving several tool calls | 30 to 50 or more credits |
Actual usage varies with the size of the input and the number of steps.
Spending less#
Match the model to the job. For an agent that mostly answers from a knowledge base, the cheapest tier of a model family costs about a third of the mid tier and usually answers just as well; the expensive tiers earn their cost on analysis and multi-step tool work. Change this in the agent's settings Model section, which Claude Agent SDK, Anthropic and OpenAI Responses agents have.
On OpenAI Responses agents, reasoning Effort is the other lever. Leave it low for lookup-style agents.
Watching usage#
The billing page (Settings → Plan) shows overall consumption and usage by model. Settings → Usage, in organizations where Runbear has rolled it out, breaks it down per agent and shows how many people used each one.
Running out#
Agents pause when the organization reaches its credit limit. Runbear emails you before that happens.
On the Free plan the included credits are the limit; upgrade to continue. On a paid plan, an Admin can let agents keep going past the included credits: on the billing page, set the Overage limit slider to the number of extra credits you allow each month and click Save. The page shows the maximum overage cost for that setting, and the credits used beyond the included amount appear as Additional credits, billed per credit. Remove limit clears the overage allowance, so agents stop again at the included credits.
Interactors#
An interactor is a person who talks to one of your agents. Runbear counts distinct interactors per billing month: the same person asking fifty questions counts once, and the count resets each cycle.
Plans#
Runbear's plans are Free, Starter, Individual, Team, Pro, Business and Enterprise. New organizations start on Free, and the billing page offers Team, Business and Enterprise, each billed Monthly or Yearly. Starter, Individual and Pro remain for existing subscriptions. See pricing for what each plan includes.
When Runbear is offering trials, an organization that has never had a paid subscription can start a 14-day trial of the Team plan from the billing page. Chat in Slack, Microsoft Teams and the web, agents and integrations all work during the trial. The public API does not: you can't create API keys, and API requests are refused until the trial ends and the paid subscription starts. Changing plans during the trial doesn't end it early. Once the paid subscription starts, existing keys work without being regenerated.
Channels by plan#
| Plan | Channels |
|---|---|
| Free, Starter, Individual | Slack, Microsoft Teams |
| Team | Slack, Microsoft Teams, Discord |
| Pro, Business, Enterprise | Slack, Microsoft Teams, Discord, HubSpot, Zendesk |
Some older Team subscriptions also include HubSpot and Zendesk. A channel your plan doesn't include shows a lock on the agent's Chatbots tab and Upgrade on Settings → Integrations; clicking either opens the plan picker. The agent's runtime can narrow this further: agents on the Claude Agent SDK runtime deploy to Slack and Microsoft Teams only. See Runtimes.
Related#
- Agent settings — choose the model and, on OpenAI Responses agents, reasoning effort
- Analytics — what your agents are being used for
- Chatbot settings — deploy an agent to a channel
- Pricing