# Limit messages and block countries > Limit how many messages one visitor can send per hour, cap one intoCHAT agent's messages per month, and block the chat in countries you choose. Your plan's monthly messages are shared by all your agents, and every visitor message uses one. The **Limits and access** card keeps one visitor, or one agent, from using up that allowance, and lets you choose countries where the chat isn't offered. The card is on the **Share** tab, below **Where this agent can appear**. To choose the websites the agent can appear on, see [Allowed domains](/en/docs/allowed-domains). ## Set the limits 1. Open your agent and go to the **Share** tab. 2. In **Limits and access**, switch on **Limit messages per visitor**, **Monthly message cap for this agent** or both, and enter a number. Add countries under **Blocked countries**. 3. Click **Save**. The card is saved on its own, apart from the rest of the agent. The new limits apply from the next message. All three are off by default. When you switch one on, it starts at a suggested value, 30 messages per hour or 1,000 replies, which you can change. ## Messages per visitor **Limit messages per visitor** sets how many messages one visitor can send this agent in an hour, from 1 to 1,000. - A visitor is counted by IP address, for this agent only. People on the same network, such as an office, share one limit. - The hour starts with the visitor's first message. It is a fixed window, not a rolling hour: a visitor who reaches the limit can write again once that hour is over. - A visitor over the limit sees "Too many messages. Please try again later." in the [widget's language](/en/docs/appearance). - A refused message doesn't use any of your messages. - Your own tests in the **Playground** aren't counted. ## Monthly message cap **Monthly message cap for this agent** sets how many replies this agent can give in one message period. - The cap counts the same way as your plan: one reply for each visitor message the agent accepts. - The period is your account's monthly message period. The count starts over when your plan's usage resets. See [Plans and limits](/en/docs/plans-and-limits). - The card shows the count, such as "120 of 1,000 this period", and the date it starts over. With the cap off it shows "120 replies this period". The agent is counted even while the cap is off, so a cap you set mid-period starts from the real figure. - Messages you send in the **Playground** count and are held to the cap, as they are for your plan. - At the cap, visitors see "This chat has reached its limit for now. Please try again later." The message doesn't mention your plan. In the **Playground** you see "The Playground couldn't reply" with the reason. - A refused message doesn't use any of your account's messages. - If you lower the cap below this period's count, the agent stops replying until the period starts over. The cap can only hold an agent below your plan, not give it more. Your plan's limit still applies to all your agents together: when the account's messages run out first, every agent stops and visitors see "Message limit reached". ## Blocked countries **Blocked countries** turns the chat off for visitors in up to 50 countries. 1. Type a country name or its two-letter code, such as `RU`, or pick it from the suggestions. You can enter several, separated by commas. 2. Press Enter or click **Add**. Each country appears as a chip; click its × to remove it. 3. Click **Save**. What a visitor in a blocked country sees: - **Script tag:** the launcher doesn't appear. The browser console shows `intoCHAT: this agent is not offered in the visitor's country (Share > Limits and access), so the chat is not shown.` - **Inline iframe and direct link:** "This chat isn't available in your region." in the widget's language. - **Chat requests:** refused with the same message. A refused request doesn't use any of your messages. The country comes from the visitor's IP address, so a visitor using a VPN gets around the block. When the country can't be told, the chat is allowed. Your **Playground** isn't affected. ## Spam and idle chats The **Spam and idle chats** card, below **Limits and access**, has two more settings. Both are off by default, and the card is saved on its own with **Save**. ### Spam protection **Spam protection** pauses a conversation that is clearly spam or abuse: - On a visitor's 2nd, 4th, 8th and 16th message in a conversation, an AI check reads it, together with your agent's name and the start of its instructions, so "unrelated to your business" is judged against what your agent is for. - It looks for advertising and promotional spam, floods of gibberish or repeated text, using the agent as a free general-purpose AI for tasks unrelated to your business, and repeated attempts to override the agent's instructions. It is meant to be cautious: a rude or angry customer, a single off-topic question, small talk or anything it isn't sure about is never flagged. - The check runs after the reply, so it never slows the chat. The reply that triggered it is given as usual. - From the visitor's next message, the chat is paused: the visitor sees "This chat has been paused." in the [widget's language](/en/docs/appearance), and the chat API refuses with the reason `spam_paused`. Refused messages don't use any of your messages. - If the check fails or takes too long, nothing is paused. - You see paused conversations with a **Spam** badge on the **Conversations** tab, and can unpause them there. See [Spam protection](/en/docs/conversations-and-dashboard#spam-protection) on that page. - Your own tests in the **Playground** and temporary chats aren't checked. The check doesn't use your plan's messages. An AI check can still be wrong. Look at the **Spam** filter on the Conversations tab now and then, and unpause any real customer. ### End idle conversations **End idle conversations** ends conversations with no message for a number of days you choose, from 1 to 30. Once a day, intoCHAT ends them and writes a short summary. If the visitor writes again, a new conversation starts, linked to the old one, and the agent gets the summary. Conversations waiting for or in a live chat aren't ended. See [Ended conversations](/en/docs/conversations-and-dashboard#ended-conversations). ## What these limits don't do - Both the visitor limit and country blocking go by IP address. Someone who changes their IP address or uses a VPN gets around them. The monthly cap holds whatever the IP address, so it is the setting that bounds an agent's usage. - They don't decide which websites can show the agent. That is the job of [allowed domains](/en/docs/allowed-domains). - intoCHAT also applies its own hourly limits to every agent. Yours come on top of them. - When an agent reaches its cap, intoCHAT doesn't send an email. The [usage alerts](/en/docs/plans-and-limits) cover your plan's messages only. A [duplicated agent](/en/docs/quick-start) keeps these settings and starts its own count at zero. If visitors see one of these messages and you don't expect it, see [Troubleshooting](/en/docs/troubleshooting).