Reference
Limits and quotas
Consolidated overview of every limit on the knowmind platform. Values as of 11 June 2026 — changes are tracked in the changelog.
| Scope | Limit | Source |
|---|---|---|
| Memories, Private plan | 2,500 | Plan matrix |
| Memories, Pro and above | unlimited | Plan matrix |
| Memory soft warning | from 80% of the plan limit (warnAt80 in stats and store responses) | Quota gate |
| API rate limit, Private plan | 30 requests per minute per token | Plan matrix |
| API rate limit, Pro plan | 60 requests per minute per token | Plan matrix |
| API rate limit, Team plan | 120 requests per minute per token | Plan matrix |
| API rate limit, Business plan | 120 requests per minute per token | Plan matrix |
| API rate limit, Enterprise | 600 requests per minute per token | Plan matrix |
| Maximum document size (upload) | 10 MB | Memory service |
| Maximum memory size (store_memory) | 100 KB | Memory service |
| Recall — number of hits (k) | 1 to 25 | MCP tool schema |
| Recall — graph hops | 0 to 3 | MCP tool schema |
| Chunk size on upload | about 500 tokens with 50 tokens overlap | Recall pipeline |
| Embedding dimension | 1024 (multilingual-e5-large) | Recall pipeline |
| Webhook body size | 1 MB | Webhook worker |
| Webhook timeout | 15 seconds | Webhook worker |
| Webhook retries | 6 attempts (immediate, 1 min, 5 min, 30 min, 2 h, 12 h) | Webhook worker |
| Dead-letter threshold | Subscription disabled after 20 consecutive failures | Webhook worker |
| Audit log retention, Private | 30 days | Plan matrix |
| Audit log retention, Pro | 90 days | Plan matrix |
| Audit log retention, Team | 12 months (365 days) | Plan matrix |
| Audit log retention, Business | 24 months (730 days) | Plan matrix |
| Audit log retention, Enterprise | 5 years (1825 days) | Plan matrix |
| OAuth code validity | 10 minutes | OAuth server |
| Magic link validity | 24 hours | Auth service |
| API token format | kmt_ + 43 base64url characters (256 bits of entropy) | Token module |
| Maximum tokens per workspace | 50 active (more on request for Enterprise) | Token module |
Behaviour on exceedance
- Rate limit (429): response carries a
Retry-Afterheader in seconds. Clients should wait and retry, ideally with jitter. - Memory limit — soft warning from 80%:
knowmind_statsand the success responses ofstore_memory/upload_documentcarry the fieldsmemoryLimit,memoriesUsedandwarnAt80plus a plain-text hint. The count covers the stored documents/memories of the workspace — NOT the auto-extracted graph nodes. - Memory limit reached (402): the write call is rejected; the response carries
plan_upgrade_requiredand indicates the required plan (REST: HTTP 402; MCP: a JSON-RPC error with the same message). Existing content and all read paths (recall, export) stay fully available. - Document too large (413): the call is rejected. Split the content and ingest in multiple calls.
- Webhook dead letter: the delivery is marked dead. After 20 consecutive failures the whole subscription is disabled and must be reactivated manually in the dashboard.