Monitor

What BetterDB-inc/monitor does

BetterDB-inc/monitor connects MCP clients to BetterDB Monitor, an observability layer for Valkey and Redis. The service retains monitoring data that is commonly transient, including slow logs, command patterns, client activity, and anomaly signals. Through MCP, an AI assistant can inspect this information while diagnosing application or database behavior.

The underlying monitor supports Valkey 8.x natively and is Redis 6+ compatible for the supported functionality. Its monitoring scope includes real-time metrics, historical latency, slow commands, hot-key activity, client attribution, cluster state, CPU and I/O thread metrics, and ACL access history. The MCP package provides 60 tools for working with that monitoring surface.

How it works

The MCP process communicates over stdio and uses BetterDB Monitor as its data source. Start it with npx @betterdb/mcp, then configure an MCP-compatible client to launch the command. Authentication uses an MCP token created under BetterDB Monitor’s Settings → MCP Tokens. Tokens are required for cloud deployments and optional for self-hosted deployments.

Queries can combine live state with retained history, which supports investigations that span a time range rather than only the current database condition. The available monitoring data can help an agent relate slow commands, traffic patterns, client identity, and resource behavior when examining a performance issue.

Setup and configuration

The MCP entry point is available through the @betterdb/mcp package:

npx @betterdb/mcp

Create an MCP token in BetterDB Monitor before connecting a client. The monitor itself can be deployed with Docker, Kubernetes and Helm, or the @betterdb/monitor CLI, but those deployment methods configure the monitoring service rather than the MCP transport. The monitor requires a Valkey or Redis instance to observe; its CLI requires Node.js 20 or newer.

The MCP server is listed for Claude Code, Cursor, and other MCP clients. Client configuration should launch the stdio command and provide the authentication details expected by the package.

Tools and capabilities

The MCP interface supports investigations such as:

• Querying current and historical database metrics.
• Reviewing slow commands and command-pattern data.
• Finding hot keys and examining access-frequency changes.
• Inspecting client activity and attributing behavior by client name and pattern.
• Exploring cluster topology, slot statistics, CPU, and key distribution.
• Examining per-thread CPU and I/O metrics available from Valkey.
• Reviewing persisted ACL access activity.
• Investigating anomalies, capacity projections, and related monitoring signals where those features are enabled in BetterDB Monitor.

BetterDB Monitor also includes monitoring for vector-search operations and AI traces, but the MCP tools expose the capabilities available through the connected monitor deployment and its enabled features.

Limitations and notes

The MCP server provides access to BetterDB Monitor rather than directly replacing the monitor’s database connection and storage configuration. Available results depend on the monitored Valkey or Redis version and on the features enabled in the BetterDB deployment. Valkey-specific capabilities, including native support for newer monitoring facilities, may not be available when connecting to Redis.

Cloud use requires an MCP token. Self-hosted use can omit the token according to the deployment configuration. The package communicates through stdio, so it is intended for MCP clients that can start local command-based servers.