text
stringlengths
0
59.1k
Each incident contains:
- **Payload**: Metric value at trigger time, threshold, and sample trace ID
- **Assignee**: Team member responsible for resolution
- **Notes**: Comments added during investigation
- **Timestamps**: Triggered at, resolved at
## Dashboard
The Alerts dashboard displays:
| Metric | Description |
| ---------------- | -------------------------------------------------- |
| Total Incidents | Number of incidents in the selected period |
| Active Incidents | Currently open, acknowledged, or snoozed incidents |
| Avg Resolve Time | Mean time from trigger to resolution |
| Sparkline | Daily incident counts over the period |
Toggle between 7-day and 30-day views using the period selector.
## Alert Evaluation
VoltOps runs a scheduled job every minute that:
1. Queries traces within each alert's time window
2. Applies the configured filters
3. Calculates the metric value
4. Compares against the threshold
5. Creates an incident if triggered and no open incident exists
6. Sends notifications respecting the cooldown period
If an incident is snoozed and the snooze period expires, the incident reopens and notifications resume.
<|endoftext|>
# source: VoltAgent__voltagent/website/observability/mcp.md type: docs
---
title: MCP
description: Use VoltOps as a read-only MCP server so AI coding agents can inspect traces and logs.
---
# VoltOps MCP
VoltOps exposes a read-only MCP server so tools like Codex, Claude Code, VS Code, Cursor, or your own VoltAgent agents can inspect real traces and logs.
This is useful when you want an AI assistant to debug production behavior with live observability data instead of guessing from code alone.
## What It Is Good For
Use VoltOps MCP when you want an assistant to:
- list the projects it can inspect
- search recent traces for a project
- open a single trace and inspect its spans
- search logs across a project
- fetch all logs attached to a specific trace
- answer debugging questions such as "why did this run fail?" or "show me the slowest traces from today"
## Available Tools
VoltOps MCP currently exposes these read-only tools:
- `voltops_list_projects`
- `voltops_search_traces`
- `voltops_get_trace`
- `voltops_search_logs`
- `voltops_get_trace_logs`
These tools are designed for debugging and investigation, not mutation. They do not create, update, or delete data.
## Access Model
VoltOps MCP uses a separate bearer token designed for observability access.
- Tokens are organization-scoped
- Tokens are read-only
- Tokens can be limited to all projects or selected projects
- Tokens can have an optional default project
- Tokens are shown once when created
- Lost tokens should be regenerated
- Access is available on the Pro plan
This is intentionally different from project secret keys. You should give assistants an MCP token, not your project secret key.
:::tip
If a token can access multiple projects and no default project is configured, the assistant should ask which project to inspect before searching traces or logs.
:::
## Create a Token
1. Open [console.voltagent.dev](https://console.voltagent.dev/).
2. Go to `Settings -> MCP`.
3. Create a token.
4. Choose the project scope, optional default project, and expiry.
5. Copy the token immediately. It will not be shown again.
## Endpoint
Hosted VoltOps endpoint:
```bash