Monthly Call Allowances

Each connector subscription includes a monthly allowance of tool calls — the number of successful actions your AI assistants can perform against that connector per billing cycle. It is a volume limit, counted per billing cycle.

It is not a speed limit, and it is not the only thing that shapes how fast your AI can work. Four limits sit side by side, and this page covers only the second one:

So having allowance left does not mean unlimited speed. The pacing and timeout limits in Connector Rate Limits still apply, and a vendor account shared with another integration has a budget StackJack does not control.

Allowances by plan tier

Three things to know about these numbers:

• They are per connector subscription, not account-wide. A workspace with HaloPSA on Pro and NinjaOne on Business has 5,000 calls for Halo tools and 50,000 for NinjaOne tools, tracked independently on each subscription's own billing cycle.
• Your portal always shows the limit that actually applies. StackJack can set a custom limit on an individual subscription (for example, an Enterprise arrangement), so treat the Dashboard and Billing pages as the source of truth for your workspace.
• Plan pricing is not listed here — current pricing is shown in the portal and at checkout.

TechTribe perk allowance

An organization that redeems the TechTribe offer receives a permanent tenant-wide 50% call-volume floor on the metered paid tiers: Pro becomes 7,500 and Business becomes 75,000 successful calls per connector cycle. The unlock applies to all existing and future connectors in that organization. Free stays 100 and Enterprise stays unlimited. An operator- configured custom allowance above the perk floor remains higher; an invalid non-positive stored limit keeps its existing fail-open behavior rather than being converted into a perk cap.

The Billing and Dashboard meters show the same effective limit the MCP gate enforces and label rows raised by the perk. Billing's separate 60-days-free checkout offer is provided through https://stackjack.io/techtribe; the call-volume unlock and Paddle billing discount are related benefits but are enforced by different shipping paths.

What counts against the allowance

• Only successful calls count. A call that fails — whether blocked by StackJack or rejected by the vendor API — is recorded for visibility but consumes nothing.
• Only that connector's tools count. Calls to halo_* tools draw from the Halo subscription's allowance and no other.
• StackJack platform tools (stackjack_*) never count. Health checks, session info, service status, and support-ticket tools are free.

Checking where you stand, before the wall

You do not have to wait for the refusal to find out how much is left.

• In the portal, the Connectors page shows a usage bar on each connected card: calls used against the limit that actually applies, the percentage, and a countdown to that connector's reset. The bar turns amber at 80% and red once the limit is reached. The Dashboard and Billing pages carry the same figures.
• From your AI assistant, the platform tool stackjack_get_quota_status returns the same picture per connector — calls used, the enforced limit, percent used, calls remaining, the reset time, and an approachingLimit flag that turns true at 80%. Like every stackjack_* tool it is free, always available, and available on every plan. Full field list: Platform Tools.

Ask for it before a large batch of work, and after any monthly_limit_exceeded refusal to see which other connectors are near their own limits.

StackJack also volunteers the warning. Once a connector passes 80%, tool results from that connector carry a short stackjack.quota note in their protocol metadata. It sits beside the tool's response, never inside it — the data your AI receives from the vendor is untouched — so whether you actually see the note depends on your AI client. stackjack_get_quota_status is the dependable way to ask.

When the allowance runs out

When a connector's counter reaches its limit, further calls to that connector's tools return a structured error to your AI assistant:

• errorType is monthly_limit_exceeded
• The message states your exact usage (for example, Usage: 5000/5000) and the current billing-cycle window (start and end), and spells out that the count is successful calls since the cycle start — not the calendar month
• Your AI is expected to relay this to you rather than retry

Nothing else breaks: other connectors keep working, platform tools keep working, and the blocked attempts are visible in your Recent Activity table. stackjack_get_quota_status still answers for every connector while one of them is walled. To restore access before the reset, upgrade the connector's plan tier — see Upgrading a Connector.

When the counter resets

The counter resets when the subscription's billing cycle renews — the cycle boundary, not the first of the calendar month. The Dashboard's connector cards show a countdown to each connector's reset.

Usage counting is deliberately conservative, and StackJack reconciles the fast counter that gates your calls against the recorded call history. Because those two are reconciled on a schedule rather than in the same instant, the figure on the portal and the figure the gate is using can differ by a few calls between passes. If a number looks wrong by a wide margin rather than a few calls, contact support with the connector name and the time.

Other gates that can block calls

The allowance check is one of several gates a tool call passes through. Calls can also be refused when a subscription's payment state needs attention (subscription_past_due, subscription_paused, subscription_canceled, subscription_period_ended) or when the tool is above the subscription's plan tier (plan_upgrade_required). The full list, with what to do about each, is in Tool Errors and Troubleshooting; payment states are explained in Canceling and Payment Issues.

Automation runs are limited separately. The allowances on this page govern connector tool calls. How many automation runs your organization may have in flight at once is a different limit, set by your Agent Runner plan (or a platform allowance if you hold none), and a run that cannot start waits in a queue rather than failing. See Agent Runner plans and concurrency slots for what sets that limit, and Concurrency, slots, and the run queue for how the queue behaves.

A note for API-minded users

Responses from the MCP endpoint carry X-RateLimit-Limit-Burst and X-RateLimit-Remaining-Burst headers. These headers are informational only. No burst limit is enforced today, so don't build tooling against them.