Manage Credits and Plan Limits

Agent Limits by Plan

Free workspaces can create up to 3 agents. Pro workspaces can create up to 10 agents. Expert workspaces can create up to 10 agents. Ultra workspaces can create unlimited agents.

Your workspace Super Agent is the built-in command center for the system. Use these plan limits for the additional agents you create for roles, workflows, deployments, and team work.

If you hit the agent limit, delete agents you no longer use, consolidate overlapping roles, or upgrade before creating more production agents.

Feature Availability

Public chatbot deployments are available on Pro and higher plans. Test the agent inside the workspace before sharing a public URL, password-protected page, or sign-in required page.

Per-visitor memory for public deployments is available on Ultra and higher plans. Use it only when the chatbot should remember context separately for each signed-in visitor.

Team workspaces, the Operator CLI for team operations, and hosted Super Agent deployments require Ultra. These features are meant for shared ownership, organization tokens, always-on availability, and team-managed tools.

Scheduled jobs, Slack, Telegram, and other supported channel connections are available across plans when the connected app is supported in your workspace. Runs still use credits, so review cadence, tool access, and run history before increasing volume.

Credits and Usage

Open Account or the dashboard usage area to review available credits, recent activity, and plan status.

Manual chats, created agents, scheduled jobs, public chatbot traffic, channel requests, Operator CLI calls, Model Context Protocol (MCP) clients, and hosted Super Agent runs can all use credits when they execute work. MCP is the standard that lets AI apps discover and call tools from another product.

Longer tasks, larger files, frequent schedules, visitor traffic, and tool-heavy workflows can consume more credits than short chat requests.

If your workspace connects ChatGPT for eligible Super Agent runs, confirm expected billing behavior with a small test and review the trace, the step-by-step run log, when usage looks different from expected.

Avoid Surprise Runs

Review scheduled jobs, public deployments, connected channels, external clients, and hosted agents regularly. Pause anything your team no longer uses.

For customer-facing or high-volume work, confirm who pays for usage, set the narrowest useful access mode, and monitor run history after launch.

If credits are used faster than expected, lower the schedule cadence, narrow the prompt, reduce unnecessary tool calls, disable per-visitor memory, or pause the deployment while you review plan limits.

Related Guides

Read "Deploy an Agent as a Public Chatbot" for Pro and higher public deployment setup, privacy, payment, and per-visitor memory options.

Read "Deploy an Always-On Hosted Super Agent" and "Set Up the Operator CLI for Your Team" for Ultra team operations.

Read "Schedule Recurring Tasks and Automations" before increasing the frequency of recurring work.

Read "Read an Agent Trace" when usage or credit behavior does not match what you expected.