# Rate limits Contexory limits how often one IP address can call the plugin API and the sign-in pages. Normal use stays far below the limit; this page is for when it does not. Source: ## The limit One limit applies. It counts requests in a fixed window of 60 seconds. Going over it gets a `429` until the window ends. The limit is the default on Contexory's hosted service. | Limit | Counted against | Default | | --- | --- | --- | | Per IP address | Every request from one IP address to the plugin API, the rest of the public API and the sign-in pages. They share one count, so a request to any of them uses the same budget. | 300 requests per minute | The limit is per IP address, not per token. Machines behind one shared address (an office network, a CI runner pool) share one budget. ## When you are limited A limited request is not processed. It gets status `429`, answered before it reaches Contexory. That response carries **no** `Retry-After` and no `RateLimit` headers, and responses within the limit do not report how much of it is left: ``` HTTP/1.1 429 Too Many Requests x-vercel-mitigated: deny ``` _A limited response_ Because the response gives no wait time, back off yourself. Waiting until the next minute is enough for the window to reset; if you keep getting 429s, use exponential backoff with jitter. A `401` is not a limit: the token is wrong, and retrying will not fix it. ## What the plugins do The Claude Code and OpenClaw plugins sync on a schedule, every 5 minutes by default. A sync that gets a 429 fails for that run: the plugin logs `sync iteration failed — Too many requests`, leaves the skills already on disk as they are, and tries again at the next scheduled sync. It does not retry sooner. One install syncing every five minutes is far below the limit. Hitting it usually means many installs behind one IP address, or a script calling the API in a loop. ## Staying under the limit Four habits keep a client well under the limit: - **Spread requests out.** The limit is per IP address, so a burst from one machine or network uses the whole budget at once. - **Leave the sync interval alone.** Shortening it multiplies requests and does not make a skill arrive any sooner than it is published. - **Stop on a 401.** Fix or replace the token. Retrying only spends your IP address's budget. - **Back off on a 429.** Wait until the next minute, then retry with exponential backoff and jitter. Do not retry in a tight loop. ## Asking for a higher limit If your use is legitimate and needs more than this limit, email [support@contexory.com](mailto:support@contexory.com) with what is calling the API, how often, and from how many IP addresses. We will look at it and tell you what we can change.