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

# Review and validation

> What happens after you add a TokenChange endpoint: operator review, automated validation checks, availability states and how endpoints recover.

Every endpoint goes through two gates before it serves traffic: an operator review, then automated validation. After that, TokenChange keeps checking it while it's live.

## Status

An endpoint's **status** shows where it is in review and whether you've paused it.

| Status | Meaning |
| - | - |
| **Awaiting approval** | Waiting for an operator to review it |
| **Rejected** | The operator rejected it. The reason appears on the endpoint. Edit the endpoint to resubmit it. |
| **Active** | Approved. It's validated and routed according to its availability. |
| **Paused** | You paused it. It receives no traffic. |
| **Suspended** | Your provider account is suspended |
| **Deleted** | Removed |

## Availability

Once approved, an endpoint's **availability** shows whether it's serving traffic.

| Availability | Meaning | Receives traffic |
| - | - | - |
| **Pending validation** | Waiting to be tested | No |
| **Validating** | A check is running | No |
| **Up** | Serving | Yes |
| **Degraded** | Recently failing requests | Only when no endpoint that's up can serve the request |
| **Down** | Out of rotation until a check passes again | No |
| **Rejected** | Failed validation | No |

### How availability changes

* A new or edited endpoint starts as **Pending validation** and is tested shortly after approval.
* If every required check passes, it becomes **Up**. If a check fails, or the endpoint stays unreachable over repeated attempts, it becomes **Rejected**. Fix the problem and edit the endpoint to try again.
* An endpoint that's up becomes **Degraded** after repeated failed requests. It returns to **Up** when test traffic succeeds.
* An endpoint that stays degraded for more than an hour, or fails a periodic re-check, becomes **Down**. It returns to **Up** when a re-check passes.
* Changing the URL, bearer token or upstream model name sends the endpoint back to **Pending validation**, because earlier results no longer apply.

Turn on **Endpoint alerts** in **Settings** to get emails when your endpoints go up, down, degraded or rejected.

## Validation checks

The required check tests that your endpoint speaks the OpenAI protocol the way TokenChange needs:

1. `/v1/models` lists your upstream model name.
2. A short streamed chat completion with a fixed arithmetic question returns the correct answer in `delta.content`. Reasoning fields are ignored.
3. The final usage chunk reports `prompt_tokens`, `completion_tokens` and `prompt_tokens_details.cached_tokens`.

Checks send as many requests at once as your endpoint has slots.

## Read validation results

Open an endpoint from **Provider** > **Endpoints**:

* **Current results** shows the latest run of each check for the current configuration.
* **Validation runs** lists every run, most recent first, with its result, trigger and when it finished.

A run's result is **Passed**, **Failed**, **Unreachable**, **Timed out** or **Skipped**. Its trigger is **Admission** for the first test, **Re-check** for periodic checks and **Manual** when an operator started it.

You can read the same runs with [`GET /v1/provider/endpoints/{id}/validations`](https://docs.tokenchange.io/api-reference/provider/list-validation-runs).


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