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

# Probe health

> Check which AI platforms DevTune probed successfully for a project, when each last answered, and any open platform outage, before trusting or retrying a Visibility read that looks stale.

Probe health tells you, per AI platform, whether DevTune's latest scheduled run for your project collected answers. Read it with `GET /projects/{projectId}/visibility/probe-health` or the `devtune_get_probe_health` MCP tool; both need `visibility.read` and return the same payload.

If Visibility numbers for a platform look stale, check probe health before retrying the read. `latestRun` shows what DevTune's latest scheduled run collected on each platform, and `lastSucceededAt` when the platform last returned an answer for any of your prompts.

## What each platform returns

The response lists each platform scheduled probes run on for the project, in display order. A paused project lists none.

| Field | Meaning |
| - | - |
| `platform` | The AI platform, such as `chatgpt-search` |
| `status` | One of the values in the table below |
| `scheduleIntervalDays` | Days between scheduled runs on the platform for your project |
| `latestRun` | The latest scheduled run in the last 30 days that has started collecting prompts: `runDate`, `finished`, and prompt counts `succeeded`, `failed`, `rateLimited` and `pending`. `null` when there is none |
| `queuedRunDate` | The day of a newer scheduled run that has not started collecting prompts yet. `latestRun` shows the run before it. `null` when there is none |
| `lastSucceededAt` | The newest prompt success on the platform in the last 30 days, from any run, including runs you started. `null` when there is none |
| `openOutages` | Platform outages DevTune published that have not ended and cover this platform, however long ago they started, with the same title, description and source link the timeline shows |

`rateLimited` counts prompts that stopped on a rate limit after every retry. `failed` counts every other prompt with no answer, including prompts still uncollected when the run finished, whatever their last attempt. `pending` counts prompts DevTune is still collecting or will retry; it is always 0 once `finished` is `true`.

The response is a snapshot. While a run is in progress, `pending` falls and the other counts rise between reads.

## Status

The first row that matches decides the status.

| `status` | When |
| - | - |
| `outage` | The rows below would read `failing` or `degraded`, and an open outage DevTune published covers the platform and was open at any time on the latest run's `runDate`, as a whole UTC day |
| `unmeasured` | There is no `latestRun`, its `runDate` is more than two schedule intervals ago, or the run is still in progress and none of its prompts has finished |
| `failing` | None of the finished prompts succeeded, including a finished run that collected no prompts |
| `degraded` | Some prompts succeeded, and some failed or were rate limited |
| `healthy` | Every finished prompt succeeded |

`openOutages` lists every open outage that covers the platform, including ones that do not set the status.

A prompt has finished once it has `succeeded`, `failed` or `rateLimited`; `pending` prompts do not count toward the status.


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