DocsLimits and credits

Limits and credits

Everything the server enforces, and why. These are checked server-side: they are not instructions to the model.

Plan gating

PlanMCPBehavior
FreeNoget_account answers with an upgrade link; other tools return a plan error.
StarterNoSame. Starter does not include API or MCP access.
ProYes60 calls per minute, 5,000 per day.
UltraYes180 calls per minute, 20,000 per day.
Grandfathered lifetime plans
Pro and Ultra lifetime accounts keep MCP access on the same terms as their monthly equivalents. Starter lifetime does not include it.

Rate limits

Limits are per workspace, not per key, and they are shared across every connected client. A limited call returns a tool error with code 4029 that names the ceiling, so the assistant can back off instead of hammering.

Sending guards

These sit on top of the rate limit and exist so an agent loop cannot burn a sending domain.

  • create_campaign: 10 per hour.
  • send_reply: 30 per hour.
  • The two budgets are independent: exhausting one does not block the other.
  • delete_list refuses to run without confirm: true.

Per-call ceilings

ToolCeilingWhy
search_b2b_leads100 rowsKeeps one call from flooding the context and the credit meter.
get_list_leads, scrape_job_leads200 rowsPage with offset for more.
export_list_csv1,000 rowsLarger exports belong in the dashboard.
add_leads_to_list500 rowsSplit larger batches; the error says so.
get_inbox50 messagesTriage in pages rather than dumping the mailbox.
suppression (add)200 addressesBulk suppression belongs in the app.

Credit costs

MCP has no separate currency. Calls spend the same monthly allowances the dashboard spends, so connecting an assistant does not add a line to your bill.

ToolSpendsHow much
search_b2b_leadsB2B credits1 per row returned
find_email, enrich_leadScrape credits1 per call
scrape_leads_startScrape credits1 per lead found, trimmed to what is left
create_campaignEmail allowance1 per email as the worker sends it
send_replyEmail allowance1 per reply
Everything elseNothingReads and state changes are free
Check before you spend
get_account is free and returns remaining credits per bucket. A well-behaved agent calls it first and sizes the work to what is left.

Audit trail

Every call is recorded with the tool, the client name, the status, credits spent and the duration. Read it in Settings, MCP, or pull it from GET /api/mcp/audit with your normal session. Rolled-up daily counts power the usage chart.

Something unclear? Ask in the app, or read the product overview.