Convex · Rate Limits

Convex Rate Limits

Convex limits are expressed primarily as concurrency (queries, mutations, actions, scheduled jobs) and per-function execution constraints rather than per-second request rates. Concurrency capacity is bound to the deployment instance class (S16, S256, D1024). Plan-level monthly quotas (function calls, GB-hours of action compute, etc.) are documented in the plans artifact and act as soft quotas with pay-as-you-go overage rather than hard throttles.

Convex Rate Limits is the machine-readable rate-limit profile for Convex on the APIs.io network, conforming to the API Commons Rate Limits specification.

It captures 28 rate-limit definitions, measuring concurrent_requests, concurrent_connections, execution_seconds, documents, and bytes.

The profile also includes 5 backoff/retry policies defined and response codes documented for throttled, quotaExceeded, and serviceUnavailable.

Tagged areas include Backend, Database, Functions, Real-Time, and Reactive.

28 Limits Throttle: 429 Quota: 429
BackendDatabaseFunctionsReal-TimeReactiveServerlessTypeScriptRate LimitingQuotasThrottling

Limits

Concurrent queries (S16 instance) deployment
concurrent_requests
16
Concurrent query executions on the S16 (Starter) instance class.
Concurrent mutations (S16 instance) deployment
concurrent_requests
16
Concurrent mutation executions on the S16 (Starter) instance class.
Concurrent queries (S256 instance) deployment
concurrent_requests
256
Concurrent query executions on the S256 (Professional) instance class.
Concurrent mutations (S256 instance) deployment
concurrent_requests
256
Concurrent mutation executions on the S256 (Professional) instance class.
Concurrent queries (D1024 instance) deployment
concurrent_requests
1024
Concurrent query executions on the D1024 (Dedicated) instance class.
Concurrent mutations (D1024 instance) deployment
concurrent_requests
512
Concurrent mutation executions on the D1024 (Dedicated) instance class.
Concurrent Convex/HTTP actions (S16) deployment
concurrent_requests
64
Concurrent Convex/HTTP actions (S256) deployment
concurrent_requests
512
Concurrent Convex/HTTP actions (D1024) deployment
concurrent_requests
1024
Concurrent Node actions (S16) deployment
concurrent_requests
64
Concurrent Node actions (S256) deployment
concurrent_requests
256
Concurrent Node actions (D1024) deployment
concurrent_requests
1024
Concurrent scheduled jobs (S16) deployment
concurrent_requests
8
Concurrent scheduled jobs (S256) deployment
concurrent_requests
256
Concurrent scheduled jobs (D1024) deployment
concurrent_requests
512
Concurrent client sessions (S16) deployment
concurrent_connections
1000
Concurrent client sessions (S256) deployment
concurrent_connections
10000
Concurrent client sessions (D1024) deployment
concurrent_connections
100000
Query/mutation user-code execution time function
execution_seconds · second
1
User code in a single query or mutation has 1 second of execution time.
Action execution duration function
execution_seconds · second
600
Convex/HTTP and Node actions may run up to 10 minutes (600 s).
Documents scanned per query/mutation function
documents
32000
Documents written per mutation function
documents
16000
Function argument size function
bytes
16777216
16 MiB max argument size; Node actions are limited to 5 MiB.
Function return value size function
bytes
16777216
HTTP response size function
bytes
20971520
20 MiB maximum HTTP action response.
Code bundle size per deployment deployment
bytes
33554432
32 MiB maximum total deployed code size.
Functions scheduled per mutation function
scheduled_functions
1000
Outstanding scheduled functions deployment
scheduled_functions
1000000

Policies

Concurrency, not RPS
Convex throttles primarily on concurrent function executions and connections per deployment instance class rather than fixed requests-per-second. Scale up by upgrading the instance class or upgrading the plan.
Soft monthly quotas
Plan quotas (function calls, action compute, storage, egress) are soft. Usage above the included amount is billed at the published per-unit overage rate rather than throttled.
Action timeout
Long-running work belongs in actions (10-minute ceiling), not queries or mutations (1-second user-code limit).
Backoff Strategy
Clients should implement exponential backoff with jitter on 429 and 5xx responses; the Convex client SDKs do this automatically.
Fair Use
Even on Business/Enterprise contracts, sustained usage that materially impacts shared infrastructure may be subject to fair-use throttling.

Sources