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

# RabbitMQ, NATS and Vault runtime evidence

> Explicit local collection of RabbitMQ node alarms, NATS JetStream consumer counters and Vault initialization and seal state from a selected Pod endpoint.

When Radar runs locally, MCP clients can use `collect_application_evidence` to read one Pod endpoint after operator approval:

| Integration | Endpoint | Evidence |
| - | - | - |
| RabbitMQ | Port 15692, `/metrics` | Node-reported disk and memory alarms |
| NATS | Port 8222, `/jsz` | JetStream pending, acknowledgement-pending and redelivery counts |
| Vault | Port 8200, `/v1/sys/health` | HTTP status and initialization, seal and standby state |

Requires Pod-read/port-forward permissions, a recognized application image and its declared default plaintext port. RabbitMQ metrics and NATS monitoring must already be enabled. Credentials, TLS, custom ports, authenticated/remote/hosted/in-cluster deployments and built-in investigations are unsupported.

In local resource details, use **Collect RabbitMQ evidence**, **Collect NATS evidence** or **Collect Vault evidence** to select a Pod and collect explicitly after a permission check. **Retry permission check** does not contact the application; neither does opening or refreshing a resource. Results show the target, observation time and selected-Pod scope. **Collection limits** expands the qualifications; **Copy observation** always includes them. If a Pod is missing from discovery, open it directly. Resource or cluster changes clear results; a replaced Pod requires a new selection and explicit collection.

These are selected-endpoint observations, not automatic Issues or an application health verdict. See [MCP collection and result semantics](https://radarhq.io/docs/features/mcp.md#explicit-local-runtime-evidence).


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