> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mka1.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Service levels

> The availability and latency objectives for MKA1 API services, and the monthly report your organization gets against them.

MKA1 publishes service level objectives (SLOs) for the API services below. After each month ends, your organization gets a report showing whether each service it used met its objectives.

## Objectives

| Service | API path | Availability | p95 latency |
| - | - | - | - |
| LLM API | `/api/v1/llm` | 99.5% | No objective |
| Authorization API | `/api/v1/authorization` | 99.9% | 500 ms |
| Agents API | `/api/v1/agents` | 99.5% | No objective |
| Sandbox API | `/api/v1/sandbox` | 99.5% | No objective |
| Compute API | `/api/v1/compute` | 99.5% | No objective |

Services that stream output or run long jobs have no latency objective. Their request duration reflects the work you asked for, not how responsive the service is.

## How service levels are measured

Every request through the MKA1 API gateway is recorded with its organization, service, status code, and duration.

* **Availability** is the share of your organization's requests to a service that did not get a `5xx` server error. Client errors such as `400`, `401`, and `429` count as available.
* **p95 latency** is the 95th-percentile request duration measured at the gateway, from receiving the request to sending the last byte of the response.
* Months are calendar months in UTC. A report is generated about an hour after the month ends.
* A report covers only the services above that your organization called that month. If your organization made no requests to them, there is no report for that month. Requests that are not tied to an organization are excluded.

The targets that applied when a report was generated are stored with it, so a later change to an objective does not alter past reports.

## View your monthly report

Organization owners and admins can view their organization's reports in the console:

1. Open **Audit**.
2. Select the **Service levels** tab.
3. Choose a month. Each service shows its request count, measured availability and p95 latency, the target for each, and whether it was **met** or **missed**.
4. Click **Download CSV** to export the month's report.

Cluster admins choose an organization from the organization selector first. Reports are per organization, so the all-organizations view has none.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.