AI Assistant & MCP
AI budgets and rate limits
How daily budgets, per-tool rate limits, and the audit log keep AI usage predictable and safe.
Per-user daily budgets
Every user in your workspace has a daily AI spending budget that resets at midnight UTC. The free tier is $1.00 per user per day, which covers a few hundred typical questions. The paid tier is $5.00 per user per day and is enabled per-workspace in Admin → Settings → Billing. Owners can override any individual user's budget up or down in Admin → Settings → Team.
Per-tool rate limits
Independent of cost, every AI tool call is rate-limited by tool tier to prevent runaway loops:
- Cheap tools (list orders, fetch a tracking number):
300 calls / minute / user - Normal tools (create a return label, update inventory):
60 calls / minute / user - Expensive tools (bulk reprice, deep marketplace push):
10 calls / minute / user
What happens when budget is exceeded
When a user crosses their daily budget, Ask ShipWave switches to read-only mode: questions still work, but any tool that performs a write action returns a friendly "budget exhausted" error. MCP / API key traffic returns HTTP
429 with a Retry-After header set to the next budget reset. Cron jobs and scheduled rules are not subject to per-user budgets — they use a separate workspace-level budget.What happens when rate-limited
Rate-limited calls return HTTP
429 with a Retry-After: header indicating when the bucket will have capacity. Ask ShipWave automatically waits and retries up to twice; MCP clients should respect the header. If you are hitting limits regularly, it is almost always a sign of a runaway loop — investigate before raising the limit.The AI Audit Log
Every AI tool call is logged in Admin → Settings → AI Audit Log, regardless of whether it came from Ask ShipWave, MCP, or a scheduled rule. Each row shows: the user, the tool name, the arguments, the result, the cost in fractions of a cent, the latency, and the source (chat / MCP / cron). You can filter by user, tool, status, and date range, and export to CSV for compliance review.
Raising limits
If your team genuinely needs more headroom, contact support@shipwave.app with usage data from the audit log. Most workspaces never hit the defaults; if you do, it is usually worth a 10-minute call to make sure the underlying workflow is healthy before raising limits.
FAQs
Does Ask ShipWave count against the same budget as MCP?
Yes. Both draw from the same per-user daily budget. A call costs the same regardless of source.
How much does a typical question cost?
Read-only questions like "show me late orders" usually cost $0.001 to $0.01. Multi-step actions ("create return labels for these 12 orders") cost more, typically $0.05 to $0.20. The audit log shows the exact cost per call.
Can I set a hard per-user cap?
Yes. In Admin → Settings → Team, set "Daily AI cap" on any user. Setting it to $0 effectively disables AI for that user. See <a href="/help/ai-assistant-overview">Ask ShipWave overview</a> for the access toggle.
What counts as "cheap" vs. "expensive"?
The tier is set per tool by ShipWave and shown on each row in the audit log. As a rule of thumb: reads are cheap, single-record writes are normal, bulk and cross-system writes are expensive.
Are limits per user or per workspace?
Budgets and rate limits are per user. Cron-triggered AI work (e.g. the <a href="/help/rule-recommender">Rule Recommender</a>) uses a separate workspace-level budget that does not affect human users.
More in AI Assistant & MCP
- Ask ShipWave: the in-dashboard AI assistant
Run natural-language queries against your orders, shipments, returns, and inventory without leaving the dashboard.
- Connecting ShipWave to Claude Desktop and Cursor
Use ShipWave as an MCP server so your local AI agent can run shipping operations directly.
- API scopes explained
Lock down what each API key and agent can do with ShipWave's granular scope system.
- The Rule Recommender
Nightly AI analyst that watches your shipping patterns and surfaces automation rule suggestions you can accept in one click.