Besides your plan's monthly answers (see Plans, answers and packs), tufu has a few safety limits. They exist so one person, a script or a runaway bot can't use up your answers, or our AI, in an afternoon. Normal traffic never comes close. The visitor and team limits count replies, each reply your bot writes, over a few hours instead of a month; the hourly limit counts answers, each reply at its weight.
When a limit is reached, visitors are never told about plans or limits: they're offered a person from your team instead.
The limits#
| Limit | How much | Applies to |
|---|---|---|
| Messages per visitor | 20 a minute | each visitor, on each bot (chat and Ask console) |
| Answers per visitor | 40 every 3 hours | each visitor, on each bot. Your signed-in team (Preview) isn't counted |
| Requests for a person | 5 an hour | each visitor, on each bot |
| Answers per team | Free: 100 every 3 hours | every answer, on tufu's AI or your own key |
| Team: 600 every 3 hours | answers on tufu's AI | |
| Team: 2,000 every 3 hours | answers on your own AI key | |
| Answers per team in an hour | 1,000 in each clock hour | answers on tufu's AI, each reply counting its weight (a stronger model or more reasoning counts more). Admins get one email when it's reached; bots pick up again when the hour is up |
| Support emails | 10 an hour per sender, 300 an hour per bot | Support by email |
| Emails to one visitor | 10 a day | your team's replies emailed to a visitor |
| Teams per person | admin of up to 3 teams | each team has its own monthly answers |
The 3-hour limits roll: each answer counts for 3 hours after it was given, so there's no reset moment to wait for.
Answers on your own AI key (see Bring your own model) still count toward the per-visitor limits and the team limit, to protect your key: on Free they share one window with answers on tufu's AI, and on Team they have their own, larger one. If tufu's AI is ever paused for everyone because usage is far above normal, bots on your own key keep answering.
What visitors see#
- Support chat: a short note that they've asked a lot in a short time, with the option to talk to a person. If the team limit or the month's answers are reached, the chat says it can't answer right now and points them to Talk to a human.
- Ask console: "That's a lot of questions at once" for the per-visitor limits; "This assistant has reached its usage limit for now" when the team is out of answers. Repeat questions are still answered from the cache.
- Slack: the bot replies in the thread that it's taking a short break, or that your team's answers are used up.
- Email: nothing changes for the customer. Their email goes to your inbox for a person to answer, and they get the usual "someone will get back to you" reply.
What your team sees#
- Chats a limit sent to a person arrive in your Inbox like any handoff.
- In Preview, your bot tells you when it's paused.
- Admins get an email at 80% of the month's answers and when they run out (Plans).
Raising a limit#
- More answers: on Team, buy an answer pack under Plan and usage, turn on auto-recharge so a pack is added before you run out, or switch a bot to your own AI key. The team limit still applies, whatever your balance.
- A higher visitor, team or teams-per-person limit: email us. If you're running a busy site or many brands, we're happy to raise them for your team.