Skip to main content

Rate Limits

CleverThis applies rate limits at multiple levels: workspace-level limits set by the plan tier, namespace-level limits configured by workspace administrators, and provider-level limits enforced by upstream LLM providers.

When a rate limit is hit, CleverThis returns 429 Too Many Requests with a Retry-After header.

Plan tier limits

These are the default workspace-level limits per plan. Administrators can request limit increases.

LimitStarterProEnterprise
Requests per minute (all namespaces)60600Custom
Tokens per minute (all namespaces)100,0001,000,000Custom
Namespaces320Unlimited
API keys10100Unlimited
Actor deployments550Unlimited

Namespace-level limits

Workspace administrators can set per-endpoint rate limits that are more restrictive than the workspace limit. For example, a dev namespace might be capped at 10 RPM to prevent runaway test scripts from consuming budget. See Managing Endpoints for how to configure these.

Provider-level limits

Each upstream LLM provider (OpenAI, Anthropic, etc.) has its own rate limits. CleverThis proxies these limits transparently — if an upstream provider returns a rate limit error, CleverThis returns a 429 to the caller with an upstream_rate_limited error code. These limits are determined by your provider account tier, not by CleverThis.

Rate limit headers

All responses from the inference API include rate limit headers:

HeaderDescription
X-RateLimit-Limit-RequestsMax requests per minute
X-RateLimit-Remaining-RequestsRemaining requests in current window
X-RateLimit-Limit-TokensMax tokens per minute
X-RateLimit-Remaining-TokensRemaining tokens in current window
X-RateLimit-Reset-RequestsTimestamp when request limit resets (Unix epoch)
Coming soon

Burst allowances (short-lived spikes above the steady-state limit) and rate limit configuration via the API are being added.