Skip to content

Limits and quotas

Every world’s limits are set by its plan. This page explains how those limits behave; for the per-plan figures, see the plan comparison.

AI tools connected over MCP are rate-limited per user, per world, per minute:

PlanRequests / minute
Basic30
Standard120
Team 5 / 10 / 20180

The cap is per person, so a team’s combined throughput scales with how many members are working in the world at once. If an agent exceeds the rate, requests come back as 429 (rate limited) — wait a moment and continue. A burst of rapid tool calls is the usual cause; if you hit it routinely, a higher plan raises the ceiling.

Each world also has caps on how much it can hold — entities, media storage, API keys (per user), and team size. The numbers per plan are in the plan comparison. A couple of behaviors worth knowing:

  • At a cap, you can’t add more of that thing until you free some up or upgrade. For example, once a world reaches its entity cap, creating new entities fails until you delete some or move to a larger plan.
  • Basic worlds have no media storage and no API keys — connect Basic worlds by signing in rather than with a key, and add media by upgrading.
  • Downgrades are blocked while you’re over the target plan’s limits. If a world has more collaborators, entities, or media than the lower plan allows, reduce usage first. See Upgrade or downgrade.

Plan and usage live in a world’s Billing and Settings in the portal. For limits a connected agent runs into mid-task, the error it receives explains which limit was hit — see connection troubleshooting.