BetterDB is a monitoring, debugging, and auditing tool for Valkey and Redis, with its core positioning being "the Valkey and Redis monitor that remembers everything." It does not just show the current state; it also persistently stores slow logs, command history, and client analytics data, allowing users to trace back what happened in the database after the fact.
Main Capabilities
- Persistent historical analysis: You can query "what happened at 3 a.m.," rather than only seeing the current state.
- Slow logs and COMMANDLOG: Pattern analysis of persisted slow queries and complete command history; Valkey 8.1+'s COMMANDLOG can capture every command, not just the slowest ones.
- Client analytics: See which clients are consuming resources, and detect abnormal buffer sizes and connection spikes.
- Anomaly detection: Automatically detect abnormal patterns in memory, CPU, and connections, and provide cluster safety protection covering split-brain, stuck replicas, and data loss.
- Cluster visualization: Interactive topology diagrams, slot heatmaps, and migration tracking.
- ACL audit trail: Track who accessed what, for compliance and debugging.
- Latency monitoring and thread metrics: P50/P95/P99 latency history by event; per-thread I/O metrics and CPU utilization charts, including I/O thread breakdown for Valkey 8.x.
- Hot key tracking: List the top 50 keys by access frequency, and show how rankings change over time.
- Key size distribution: Built from the server-side INFO keysizes histogram, showing element counts for collections and byte lengths for strings, separated by database, with no need to scan keys.
- MONITOR capture sessions: Capture MONITOR streams on demand, with real-time following, filtering, and value redaction, and the ability to trigger on anomalies, schedule with cron, compare two captures, and export JSON or CSV.
- Prometheus and OpenTelemetry: Provides 100+ Prometheus metrics, supports OTLP ingestion and distribution, and Webhook alerts.
- Vector workload analysis: Operations per second and latency trends for FT.SEARCH, health checks by index, and alerts when index failures, backfill progress, or deleted document growth occur; you can inspect vector indexes, browse embedding vectors, and find semantically similar entries.
- Inference pipeline latency: P50/P95/P99 distribution by operation for FT.SEARCH, reads, and writes, with SLA thresholds set per index and Webhook alerts when breached.
- Throughput forecasting: Operations per second forecasts based on growth rate trends and upper limits, with alerts before capacity is reached.
- Cross-database compatible migration: Analyze compatibility, migrate data, and verify results, supporting migration in any direction among Redis, Valkey, cloud, and self-hosted.
How to Use
Self-hosting is supported, starting with a single command, then visit http://localhost:3001，可参考 Docker guide. Cloud deployment is also available, with no infrastructure to maintain. The product can run fully offline in air-gapped environments; offline license tokens are verified locally, with no internet connection required, and data does not leave the user's network.
For agent scenarios, BetterDB provides an MCP server that can connect to Claude Code, Cursor, or any MCP-compatible client, allowing users to ask questions in natural language and get structured diagnostic results, such as slow log history, anomaly summaries, latency trends, ACL audits, and more, without opening a browser.
Applicable Scenarios and Limitations
Suitable for teams running Valkey or Redis, especially scenarios requiring persistent observability, compliance auditing, cluster safety inspections, and production monitoring of vector retrieval (RAG and semantic search). It is officially described as having an open-source core and zero lock-in, and is currently in early access, with all Pro and Enterprise features available for free during the beta period. Facts such as specific pricing and official release timing are not yet provided in the materials.
