MCP Redis Diagnostics

MCP server for Redis diagnostics — analyze memory usage, slowlog, client connections, and keyspace health with AI-powered recommendations.

Why This Tool?

Most Redis MCP servers are CRUD wrappers (get/set keys). RedisNexus offers diagnostics but targets enterprises (K8s, multi-tenant SaaS). This tool is the only lightweight npm package for deep Redis diagnostics — 7 tools covering memory fragmentation, slowlog patterns, client connection health, keyspace distribution, latency analysis, and configuration auditing. Install with npx, no Docker or SaaS required.

Related MCP server: mcp-redis-server

Pro Tier

Generate exportable diagnostic reports (HTML + PDF) with a Pro license key.

• Full JVM thread dump analysis report with actionable recommendations

• PDF export for sharing with your team

• Priority support

$9.00/month — Get Pro License

Pro license key activates the generate_report MCP tool in mcp-jvm-diagnostics.

Tools (7)

analyze_memory

Analyze Redis memory usage and fragmentation.

Detects:

• High memory fragmentation (>1.5x RSS/used ratio)

• Swap risk (fragmentation <1.0)

• Maxmemory pressure (approaching limit)

• Eviction patterns

• Missing maxmemory configuration

• Unsafe noeviction policy

analyze_slowlog

Analyze Redis SLOWLOG for slow commands.

Parameters:

• count (number, default: 128) — Number of slowlog entries to retrieve

Detects:

• Dangerous O(N) commands: KEYS, SMEMBERS, HGETALL, SORT

• High latency commands (>10ms, >100ms thresholds)

• Command concentration patterns

• Full slowlog buffer (missing history)

analyze_clients

Analyze Redis client connections.

Detects:

• Blocked clients (BLPOP/BRPOP)

• Connection pool saturation (>80% maxclients)

• Idle connections (>5 minutes)

• Large output buffer memory

• Pub/sub subscriber patterns

analyze_keyspace

Analyze Redis keyspace distribution and cache effectiveness.

Detects:

• Low TTL coverage (<20% of keys)

• Low cache hit rate (<80%)

• Unbalanced database distribution

• High expiry/eviction rates

• Multiple database anti-pattern

analyze_latency

Analyze Redis latency events from the LATENCY subsystem.

Detects:

• Fork latency spikes (RDB/AOF background save blocking operations)

• AOF fsync delays and write latency

• Slow command processing (O(1) commands unexpectedly slow)

• Eviction and key expiry cycle delays

• Active defragmentation impact

• Increasing latency trends over time

Requires latency-monitor-threshold to be set in redis.conf (e.g., CONFIG SET latency-monitor-threshold 100).

analyze_config

Analyze Redis configuration for security and reliability risks.

Detects:

• No maxmemory limit (unbounded memory growth, OOM risk)

• Unsafe eviction policy (noeviction causing errors at memory limit)

• Network exposure (bind 0.0.0.0 without protected-mode)

• Missing authentication (no requirepass)

• Disabled persistence (both AOF and RDB off — data loss on restart)

• Idle connection accumulation (timeout 0)

• Disabled TCP keepalive (dead connections undetected)

• Low server frequency (hz < 10 slowing background tasks)

analyze_performance

Comprehensive health assessment — runs all analyzers and produces a unified report.

Parameters:

• slowlog_count (number, default: 128) — Number of slowlog entries

Installation

npm install -g mcp-redis-diagnostics

Or run directly:

npx mcp-redis-diagnostics

Configuration

Environment Variables

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "redis-diagnostics": {
      "command": "npx",
      "args": ["-y", "mcp-redis-diagnostics"],
      "env": {
        "REDIS_URL": "redis://localhost:6379"
      }
    }
  }
}

Redis with password:

{
  "env": {
    "REDIS_URL": "redis://:yourpassword@localhost:6379"
  }
}

Quick Demo

Once configured, try these prompts in Claude:

1. "Analyze my Redis memory usage — is there fragmentation?" — Shows used vs max memory, fragmentation ratio, eviction policy, and memory pressure issues

2. "Check the Redis slowlog for dangerous commands" — Identifies O(N) commands like KEYS/SMEMBERS, high-latency patterns, and optimization suggestions

3. "Run a complete Redis health check" — Unified report combining memory, slowlog, clients, keyspace, and latency analysis

"What's my cache hit rate? Are my TTLs configured properly?"

"Give me a full Redis health check"

Part of the MCP Java Backend Suite

This tool is part of a suite of MCP servers for backend developers:

• mcp-db-analyzer — PostgreSQL/MySQL/SQLite schema analysis

• mcp-jvm-diagnostics — Thread dump and GC log analysis

• mcp-migration-advisor — Flyway/Liquibase migration risk analysis

• mcp-spring-boot-actuator — Spring Boot health and metrics analysis

• mcp-redis-diagnostics — Redis memory, slowlog, and client diagnostics

Limitations & Known Issues

• Single Redis instance: Analyzes one Redis instance at a time. Does not support Redis Cluster topology discovery or Sentinel failover analysis.

• ACL restrictions: Some tools require specific Redis commands (SLOWLOG, CLIENT LIST, LATENCY). Redis ACLs may block these. The analyze_performance unified tool handles partial failures gracefully.

• Latency monitoring: The analyze_latency tool requires latency-monitor-threshold to be set in redis.conf. Without it, no latency events are captured.

• Key-level analysis: Keyspace analysis uses INFO keyspace aggregates. Individual key inspection (e.g., finding the largest keys) requires MEMORY USAGE per key, which is not performed to avoid impacting production.

• Redis Cluster: No cluster-specific analysis (slot distribution, rebalancing, cross-node latency). Works against individual nodes only.

• Redis Modules: Module-specific commands and data types (RedisJSON, RediSearch, RedisTimeSeries) are not analyzed.

• Memory advisor: Memory recommendations are based on INFO memory stats. For detailed memory breakdown by key type, use redis-cli --bigkeys externally.

• Fragmentation ratio: Memory fragmentation uses RSS vs. used memory ratio, which can be distorted by jemalloc. Values <1.0 may not always indicate swapping.

• Read-only: All commands are read-only (INFO, SLOWLOG GET, CLIENT LIST, LATENCY). No data or configuration is modified.

License

MIT

End-of-life: 2026-05-10.

This MCP server is no longer maintained or distributed. The Corporation has pivoted to Apify marketplace actors. See irrationalways on Apify and irrcorp/bzp-poland-tenders for current Corporation work.

The npm package has been unpublished. The repository is archived for historical reference only.