BLog
/
/
User, Team, Client, or Tool: Where Each Rate Limit Belongs
Content
↳
Example H2
↳
Example H3
↳
Example H4
↳
Example H5

You open a user in PopdockAI's security settings and find a Rate limits tab. Then you find the same tab on a team, on an MCP client, and on an MCP tool. Then a Default rate limits tab on the Account page. Five places to set what sounds like one thing.

It's a fair question which one matters. The answer is that they all do, at the same time, and each one is the right tool for a different problem. Put a limit in the wrong place and it either does nothing useful or catches people it was never meant for.

This article walks through each place a rate limit can live, how they combine, and a sensible place to start.

How Every Rate Limit Works

Every rule, wherever you add it, is built in the same Add rate limit window from the same few fields:

  1. Applies to. MCP tool calls, made by AI clients through your MCP servers, or API calls, made to the PopdockAI API by scripts, integrations and the command line tools. The two are counted separately.
  2. Server and Tool. For MCP tool calls, you can narrow a rule to one server, or to one tool on that server. Choose All servers to cover every tool call, or a server with All tools to cover everything on that server.
  3. Period. Per minute, Per hour or Per day.
  4. Max calls. How many calls are allowed in that period.

Each user, team, client or tool can have one rule per combination of calls, tool and period, so a user can have both "60 MCP tool calls per minute" and "2,000 per day". PopdockAI refuses a second rule for the same combination and tells you to edit the existing one instead.

The Five Places, and What Each One Counts

The fields are the same everywhere. What changes is whose calls go into the count.

1. Default rate limits, on the Account page. Every user gets their own allowance. A default of 60 MCP tool calls per minute means each person gets 60, not 60 shared between everyone. This is the safety net for the whole account.

2. A user's Rate limits tab. A limit for one person, counted against their calls only. A user's own rule replaces the default for the same calls, so this is how you give a power user more room, or an automation account less. The replacement is exact: a user rule on one server leaves the all-tools default in force for everything else.

3. A team's Rate limits tab. One allowance shared by everyone on the team. A team limit of 1,000 calls per hour means the whole team's calls add up to 1,000, however they're split. A user on two teams counts against both.

4. An MCP client's Rate limits tab. One allowance for every call that comes through that client, whoever is using it. Use this to cap a whole rollout, like the AI assistant the support team uses, without touching anyone's personal limits elsewhere.

5. An MCP tool's Rate limits tab. One allowance for that tool, shared by everyone in the account. This is the limit for a tool that's slow, expensive on the source system, or writes data back. The rule stays on the tool when you publish a new version of the server, so you don't have to set it again.

MCP clients and MCP tools only ever see MCP tool calls, so their tabs don't offer API calls. API calls can be limited by user, team or default.

How They Combine

PopdockAI doesn't choose one rule per call. It checks every rule that matches and refuses the call if any one of them is used up.

Say Priya, an analyst on the finance team, asks her AI assistant a question that calls the "Get customer invoices" tool. That one call counts against:

  • Her own allowance, either her user rule or, if she has none for those calls, the account default.
  • The finance team's shared allowance, if the team has one.
  • The AI client's allowance, if the client has one.
  • The tool's account-wide allowance, if the tool has one.

Whichever one runs out first is the one that stops her, and the refusal says which: "You are limited to...", "Your team is limited to...", "This AI client is limited to..." or "This tool is limited to... across the account." That wording is deliberate. "You" and "your team" call for different fixes, and the person who hits the limit should know who to ask.

Because every rule applies, you never need to worry that adding a team limit loosens a user's limit. A rule can only make things stricter. The one exception is the default, which a user's own rule replaces for the same calls. That's how you give one person more room.

How to Decide

Match the limit to the problem you're actually worried about.

Table 1
You're worried about Put the limit on Example
Any one person's agent running away Default rate limits 60 MCP tool calls per minute
A specific person or service account That user 10 API calls per minute for a nightly sync
One department crowding out the rest The team 3,000 MCP tool calls per day
A new AI client rollout The MCP client 1,000 MCP tool calls per hour
A tool that writes back or hits a slow system The MCP tool 20 calls per minute

‍

Our recommended starting point is a per-minute default on MCP tool calls, high enough that no person typing questions will ever notice it, and low enough to catch an agent in a loop within a minute or two. Then add a tool limit to every tool that creates, updates or deletes records. Those two rules cover most of what goes wrong, and you can add team and client limits later, once you've seen how people actually use it.

Per-minute limits catch loops quickly. Per-day limits cap the total for a job that might run all afternoon. A rule for each is a sound pairing on anything you're nervous about.

Details People Get Wrong

Periods are fixed, not rolling. A per-minute limit resets at the top of each minute, a per-hour limit at the top of the hour, and a per-day limit at midnight UTC. If you're outside UTC, "per day" resets at a time of day you may not expect, so set daily limits with some headroom.

Changes take up to a minute. PopdockAI caches the rules close to where calls arrive. Lowering a limit to stop a loop that's already running takes effect within a minute, not instantly.

Refused calls are free. A refused MCP tool call doesn't spend a task from the monthly allowance, so a limit set a little low costs you a short wait, not money.

Code mode doesn't get around anything. When a code mode script calls five tools, each call counts against every matching limit, including tool limits. The script itself doesn't count on top.

Who Can Set Them

Default and user rate limits need permission to manage security. Team rate limits need permission to update teams. MCP client and MCP tool rate limits are part of PopdockAI's MCP features and are available on plans that include MCP. If you're not sure which plan your account is on, your Popdock administrator can check it on the Subscription tab.

‍

From AI Potential to AI Performance in Minutes

One layer.
All your apps.
‍Any AI tool.

Join forward-thinking companies that are transforming their operations with intelligent, secure AI automation.

Try PopdockAI Free

See Documentation