> ## Documentation Index > Fetch the complete documentation index at: https://docs.minidesk.ai/llms.txt > Use this file to discover all available pages before exploring further. # Sending limits > How fast a workspace can reply to customers, and why there is a limit at all. Every workspace has a maximum rate at which replies leave it. This is separate from [the API rate limits](/guides/rate-limits), which count requests against a credential. This counts **messages that reach a customer**, whoever sent them: an agent typing in the dashboard, an automated AI reply, or a reply posted through the API. ## Why this exists A helpdesk with an unmetered outbound channel is a spam relay waiting to be found. Someone signs up, points it at a list, and the mail goes out over the same IPs and the same domain that carry everyone else's support replies. That damage takes months to undo and about a day to cause. These are **rates, not quotas**. Nothing here caps how many conversations you handle or how many replies you send in a month. There is no allowance to run out of and no upgrade prompt at the end of it. The limit is on how fast, and it is set well above the speed a person types. ## The limits | Plan | Per minute | Per hour | Per day | New recipients per hour | | ------- | ---------- | -------- | ------- | ----------------------- | | Free | 10 | 120 | 1,000 | 15 | | Starter | 30 | 600 | 10,000 | 60 | | Growth | 60 | 1,800 | 30,000 | No limit | | Scale | 120 | 6,000 | 100,000 | No limit | All four are counted per workspace. Running three workspaces gives you three sets, the same way [plans](https://minidesk.ai/pricing) work everywhere else. The three message windows do different jobs. The minute stops a script firing everything at once, the hour stops something pacing itself through an afternoon, and the day catches a slow drip that would clear both. ## Every channel shares one budget The numbers above are **per workspace, not per channel**. Email and Telegram draw from the same 10 a minute on Free rather than 10 each, and any channel added in future joins the same pool. That is deliberate. Separate budgets would mean connecting a second channel doubled how fast a workspace could send, which is the opposite of what a limit is for. Live chat replies are not counted. A message in the chat widget goes to a visitor who has the window open in front of them, so it cannot reach anybody who did not ask for it. ## New recipients The last column is the one that matters most, and it is the reason the other three can be as generous as they are. A **new recipient** is an email address or chat account that has never written to you. Reply to somebody who contacted your support desk and it does not count against this limit, however long ago they wrote. Send to an address that has never been in touch and it does. That distinction is what separates a support desk from a bulk sender. A desk answers people who contacted it, so this counter stays near zero in normal use. You will only meet it if you are opening conversations with people who did not ask, which is the thing this limit is here to slow down. Growth and Scale have no limit on it. ## What happens when you hit one The reply is not sent, and an internal note appears on the conversation saying so and naming when sending resumes: > This reply was not sent. This workspace is sending faster than its plan > allows and is being rate limited to protect delivery for everyone. The > customer did not receive it. Sending resumes at the top of the hour; send the > reply again after that. The note is internal, so the customer never sees it. Nothing is queued and nothing is retried in the background: send the reply again once the window has cleared. A reply held by a limit **did not reach the customer**. The message is still on the conversation and still in your history, which is exactly why it is worth reading the internal note rather than assuming a sent-looking message was sent. ## Receiving There is a ceiling on how many conversations a workspace can take in as well: 500 a day on Free, rising to 150,000 on Scale. It exists so one workspace cannot flood the platform, and it is set far above what any real support desk receives. If you are near it, something is wrong rather than busy, and it is worth [getting in touch](mailto:hello@minidesk.ai) before it bites. Conversations refused this way return a `429` with a `rate_limited` code, the same shape as [any other error](/guides/errors). ## If a limit is in your way If you have a legitimate reason to send faster or to open conversations with people who have not written in, [email us](mailto:hello@minidesk.ai). These are anti-abuse numbers rather than product tiers, and we would rather raise one for a real customer than have you work around it.