## TL;DR
An expired Neptune API credential blocks new runs at creation, while existing connections may keep working briefly: issue a fresh API token from the workspace settings, update the environment or secret the job reads, and restart the job so it loads the new value. Record the new token's expiry where the team tracks rotations so the next expiry is a calendar event, not an outage.

```text
Neptune auth token expired error failed to create new experiment run
```

## Use this when
- The run or call fails with this symptom: Neptune auth token expired when creating a new experiment run.
- You are working on a Neptune job or agent creating experiment runs on a schedule and the failure repeats on the same run or configuration.
- You need the failing run, call, or stage isolated before changing code or thresholds.

## Not for this skill when
- A token rejected because it belongs to a different workspace, where the fix is minting one in the right workspace.
- The error text names a different tool, stage, or root cause. Match the verbatim error first and use that skill.

## Steps
1. Confirm the failure is authentication by initialising a minimal run with the current token and reading the exact error. Verify: the token is shown to be rejected
2. Generate a new API token in the workspace the runs belong to. Verify: a fresh token exists for the correct workspace
3. Update the job's secret or environment and restart it so the new token is loaded. Verify: the job starts with the new credential
4. Create a test run and log one value to prove the path end to end. Verify: the test run appears with the logged value

## Variant phrasings
### neptune auth token expired create run error
The verbatim phrasing.
### neptune api token expired automation rotation fix
The rotation phrasing.
### neptune run creation authentication failed token
The creation phrasing.

Compatibility: Neptune authentication for automated run creation

## Why it happens
Tokens are read once at process start and expire on the issuer's schedule, not the job's. Long-lived automation therefore fails suddenly at an ordinary run creation, and the fix is rotation plus a record of when the next rotation is due.

## Edge cases / pitfalls
- Workspace-level service tokens outlive personal ones on some plans; prefer them for shared automation.
- A token pasted with trailing whitespace reads as a different, invalid token; store it programmatically.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_iySy3kbqEGDCPWYUfJMXwg