# Cost

> Fifteen read-only MCP tools cover cloud spend: totals by provider, team or tag, daily and monthly trends, budget health, detected anomalies, and per-resource cost history. Every one is tier 0, so an assistant can answer any spend question without an org admin enabling writes at all.

Source: https://zop.dev/integrations/mcp-server/tools/cost
Updated: 2026-08-20

---

## Common questions

### Can an assistant answer cost questions with writes turned off?

Yes. All fifteen tools in this category are tier 0, so the whole group works at the default org setting. Nothing here creates, modifies or deletes anything; create_budget and update_budget deliberately live in govern behind tier 1.

### Which cost tool should an assistant call first?

get_cost_overview. It is the entry point, and it returns the org-wide picture an assistant needs before drilling into a provider, team or tag.

### How do I know a quoted figure is not stale?

get_billing_sync_status reports billing sync health per cloud account, so an assistant can say when billing last synced instead of presenting old totals as current. The figures themselves come from the same daily rollup the dashboards read.

## What an assistant can answer with these

Start with `get_cost_overview`. It is the entry point, and it returns the org-wide picture an
assistant needs before drilling anywhere. From there the tools split by the question being asked:

- **"Where is the money going?"**: `get_cost_by_provider`, `get_cost_by_resources`,
  `get_cost_by_team`, `get_cost_by_tag`. The last two are showback: they attribute spend to the
  owning team or tag rather than the account that happens to hold the resource.
- **"Is it going up?"**: `get_cost_trends` for daily, weekly or monthly movement, and
  `get_resource_cost_history` for one resource over time.
- **"Did something break?"**: `get_anomalies` returns detected spikes for a period,
  `get_billing_sync_status` tells you whether the underlying billing data is current.
- **"Are we within budget?"**: `list_budgets` and `get_budget_summary`.

`get_ai_usage` is the odd one out: it reports AI Gateway spend, one row per LLM request, so an
assistant can account for its own cost alongside your infrastructure.

## Why every tool here is read-only

Cost data is the one surface where an assistant should never need write access, and treating it
that way means the whole category works at the default org setting. Nothing in this group can
create, modify or delete anything; `create_budget` and `update_budget` deliberately live in
the `govern` category instead, behind tier 1.

## Numbers are reconciled, not estimated

These tools read the same daily billing rollup the dashboards use, so a figure an assistant
quotes matches what a human sees in the UI. `get_billing_sync_status` exists so an assistant can
say "billing last synced 4 hours ago" instead of silently reporting stale totals as current.