> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.talqora.com/operate-at-scale/usage-and-limits/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.talqora.com/_mcp/server.

# Usage and limits

Talqora bills storage, data written, vector queries, query processing, and returned data separately. The console reports logical bytes so teams can explain their usage before an invoice arrives.

* Maximum dimensions: 4096
* Maximum vectors per write or delete: 500
* Maximum request body: 20 MiB
* Writes and deletes require an `Idempotency-Key`
* Rate limits are plan-specific and aggregated by organization
* Rate-limited calls return `429`, `Retry-After`, and `RateLimit-*` headers

API-key management still uses the authenticated dashboard session. Index endpoints accept either a dashboard session or a Talqora API key: `read` lists and reads scoped indexes, while an unscoped `write` key can create indexes and a scoped `write` key can rename, branch, or delete its allowed indexes. Vector reads and writes use a Talqora API key.

One write request can contain up to 500 vectors and consumes one request from the write-rate bucket. Its rows still count individually toward the monthly write allowance. See [Rate limits](./rate-limits) for every plan's request budget and retry contract.