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

# Doctor And Health

> Diagnose and repair a broken local setup.

# Doctor And Health

Loong uses `doctor` as the main repair path when the base runtime is unhealthy.

## Start Here

| If you need to...                                  | Start with                                                  |
| -------------------------------------------------- | ----------------------------------------------------------- |
| check the local runtime health quickly             | `loong doctor`                                              |
| apply the safe repair path                         | `loong doctor --fix`                                        |
| feed health output into another tool or script     | `loong doctor --json`                                       |
| inspect the security-oriented checks only          | `loong doctor security`                                     |
| keep a live developer watch on one runtime/session | `loong debug watch --config <path> --session-id <session>`  |
| capture one deeper developer/debug bundle          | `loong debug bundle --config <path> --session-id <session>` |

## Core Commands

```bash theme={null}
loong doctor
loong doctor --fix
loong doctor --json
loong doctor security
loong doctor security --json
loong debug watch --config /path/to/loong.toml --session-id session-123
loong debug watch --config /path/to/loong.toml --session-id session-123 --refresh-ms 1000
loong debug bundle --config /path/to/loong.toml --session-id session-123
loong debug bundle --config /path/to/loong.toml --session-id session-123 --output /tmp/debug-bundle.json
loong debug show --artifact /tmp/debug-bundle.json
```

## When To Escalate To Audit

```bash theme={null}
loong audit recent --limit 20
loong audit summary --limit 200 --json
loong audit token-trail --token-id token-abc
```

Use the audit commands when:

* the runtime looks unhealthy and you need recent operator-visible evidence
* a repair action happened but you want to inspect what changed
* policy or token flow needs a more explicit trail than `doctor` alone provides

## When To Escalate To Debug Bundle

```bash theme={null}
loong debug watch --config /path/to/loong.toml --session-id session-123
loong debug bundle --config /path/to/loong.toml --session-id session-123
loong debug bundle --config /path/to/loong.toml --session-id session-123 --output /tmp/debug-bundle.json
loong debug show --artifact /tmp/debug-bundle.json
```

Use the debug bundle when:

* you need a live, continuously refreshed view over one active runtime or session before you decide whether to capture a point-in-time artifact
* the runtime is behaving incorrectly and you need one combined view over status, ACP, session, provider failover, and audit evidence
* you want to hand one artifact to another maintainer without asking them to rebuild the same local inspection steps
* you need request ids, failover rollups, session recovery state, and audit integrity in one place

## What This Page Optimizes For

* keep health issues legible for operators
* prefer safe local repair actions
* surface concrete next steps instead of vague status-only output
* preserve audit visibility for policy and runtime debugging

## When To Use It

* after onboarding if credentials or config still feel incomplete
* before enabling channels or gateway surfaces
* when runtime behavior has drifted and you need a trustworthy repair path

## Typical Order

1. Run `loong doctor`.
2. Run `loong doctor --fix` if the safe repair path is appropriate.
3. Use `loong debug watch` when you need live counters, recent session events, and failover/audit attention while the issue is still happening.
4. Capture `loong debug bundle` when you need a stable artifact to share or compare later.
5. Use audit commands when you need evidence or a deeper governance trail.
6. Return to [First Run](/get-started/first-run) once the base path is healthy again.

## What Comes Next

* Return to [First Run](/get-started/first-run) if you still have not reached a first useful answer.
* Continue to [Skills](/use-loong/skills) when the blocker is install policy, skill exposure, or browser automation setup.
* Continue to [Browser And Web Boundaries](/use-loong/browser-and-web-boundaries) when the blocker is a private host, domain allowlist, or browser runtime gate.
* Move on to [Use Loong](/use-loong/overview) once the base local runtime is healthy.