Service Levels

1. Scope

These service levels cover two services:

• The Data API at https://api.vaquill.ai, including the metered US law endpoints.
• The MCP server at https://mcp.vaquill.ai, including OAuth sign-in for connected AI clients.

They do not cover the marketing website, the open-data downloads, or features marked as beta, which can change or be withdrawn at short notice.

2. Availability target

We target 99.5% availability per calendar month for each service. That allows about three and a half hours of unplanned downtime in a 30-day month.

This is a target we publish and measure ourselves against, not a guarantee, and it carries no service credits on self-serve plans. Customers who need a contractual commitment with service credits can agree one on an enterprise order form; contact contact@vaquill.ai.

2.1 How availability is measured

An external monitor checks the health endpoint of each service at least every five minutes from outside our infrastructure. A service counts as unavailable for the time its checks fail. Monthly availability is the share of the month in which the service was available, and the results are published on our status page.

2.2 What does not count as downtime

• Scheduled maintenance announced on the status page at least 48 hours in advance.
• Requests refused by design: HTTP 429 when a rate limit is exceeded, HTTP 402 when an account is out of credits, and HTTP 400, 401, 404 or 422 for invalid credentials or requests.
• Problems outside our infrastructure, including the caller’s network, an AI client’s own outage, and events covered by the force majeure clause of the Terms of Service.
• Delays in how quickly a government publisher’s changes reach our corpus. Data currency is reported per section and per corpus in the API itself; it is not an availability measure.

2.3 Failed requests are not charged

A request that fails because of a fault on our side (HTTP 500 or 503) is refunded automatically, and every metered response reports the credits it actually consumed.

3. Support response targets

Contact support at contact@vaquill.ai or through the support page. A first response is a reply from a person who has read your request, not an automated acknowledgment.

A business day is Monday to Friday, excluding public holidays.

4. Incidents and maintenance

• During an outage or a significant degradation we post updates on the status page until it is resolved.
• Scheduled maintenance is announced there at least 48 hours in advance.
• Breaking changes to the API or the MCP tool catalogue are announced in advance in the documentation and the changelog.

5. Relationship to the Terms of Service

This page describes our service levels; the Terms of Service govern. The targets here are not a warranty and do not give rise to service credits unless a signed order form says otherwise. We may update this page, and we will not lower a target without announcing it here first.