BetterDB MCP server

Name: BetterDB MCP server
Author: BetterDB

Query a BetterDB monitor's Valkey and Redis metrics, slowlogs, hot keys and cache proposals from a chat

Official · BetterDB TypeScript Local · stdio DevOps and monitoring Databases

GitHub npm Website

43 popularity score ★ 1.3k ↓ 75/wk

Connect your client

Local (npm)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server. Then replace each <NAME> in its settings with your value.

Or add it by command or config

It needs BETTERDB_URL. Replace each <NAME> with your value.

Claude CodeTerminal

claude mcp add betterdb -e BETTERDB_URL=<BETTERDB_URL> -- npx @betterdb/mcp

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "betterdb": {
      "command": "npx",
      "args": [
        "@betterdb/mcp"
      ],
      "env": {
        "BETTERDB_URL": "<BETTERDB_URL>"
      }
    }
  }
}

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "betterdb": {
      "command": "npx",
      "args": [
        "@betterdb/mcp"
      ],
      "env": {
        "BETTERDB_URL": "<BETTERDB_URL>"
      }
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "betterdb": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "@betterdb/mcp"
      ],
      "env": {
        "BETTERDB_URL": "<BETTERDB_URL>"
      }
    }
  }
}

Codex~/.codex/config.toml

[mcp_servers.betterdb]
command = "npx"
args = ["@betterdb/mcp"]

[mcp_servers.betterdb.env]
BETTERDB_URL = "<BETTERDB_URL>"

Gemini CLITerminal

gemini mcp add --scope user -e BETTERDB_URL=<BETTERDB_URL> betterdb npx -- @betterdb/mcp

Settings 4

`BETTERDB_URL`	Required: Address of the BetterDB monitor or Cloud workspace (https://<workspace>.app.betterdb.com). Default `http://localhost:3001`.
`BETTERDB_TOKEN` secret	Optional: MCP token from Settings → MCP Tokens, sent as a Bearer header. Needed for BetterDB Cloud, optional for a self-hosted monitor.
`BETTERDB_INSTANCE_ID`	Optional: Instance to use from the start, so select_instance is not needed. For select_instance.
`BETTERDB_TELEMETRY`	Optional: Set to false to stop the tool-call usage reports the server posts to the monitor. For telemetry.

What it does

The BetterDB MCP server (npm package @betterdb/mcp) is the assistant-facing side of BetterDB Monitor, an observability tool for Valkey and Redis. It does not open a database connection itself: every tool calls the REST API of a running BetterDB monitor, either a self-hosted one (http://localhost:3001 by default) or a BetterDB Cloud workspace, and returns what that monitor has collected.

Through it an assistant can pick one of the monitored instances and read its health summary, INFO sections, slowlog and COMMANDLOG entries (live and persisted, with pattern aggregation), latency events, memory diagnostics, client lists and activity over time, hot and largest keys, the ACL audit log, cluster node and per-slot stats, capacity forecasts, anomaly events, vector index health and FT.SEARCH latency, and OpenTelemetry traces from AI applications correlated with Valkey state.

It can also change things: add, test, remove or set the default monitored connection; start and stop a local monitor process; and work with the review flow for BetterDB's semantic and agent caches and agent-memory stores, where tools propose a threshold, TTL, invalidation or forget, and separate approve tools apply the change to Valkey.

The version 1.5.0 source registers 60 tools (packages/mcp/src/index.ts at tag mcp-v1.5.0). The package README lists 51 of them: it leaves out the nine agent-memory tools (memory_stores, memory_list, memory_get, memory_stats, memory_recall, memory_forget, memory_list_pending_forgets, memory_approve_forget, memory_reject_forget). With --autostart the server launches the monitor itself by running npx --yes @betterdb/monitor, records the process id in ~/.betterdb/monitor.pid, and gives up if the monitor is not healthy within 15 seconds (src/autostart.ts).

What it needs

Runtime	Node.js 20+
Runs on	macOS, Windows, Linux
BetterDB	Your own instanceA running BetterDB monitor (self-hosted or BetterDB Cloud) that watches a Valkey 8.0+ or Redis 6+ instance. Cloud needs an MCP token.
Also	npx on PATH, also used by --autostart to fetch @betterdb/monitor

On BetterDB Cloud, create a token under Settings → MCP Tokens and set it as BETTERDB_TOKEN. A self-hosted monitor can run from the betterdb/monitor Docker image, npx @betterdb/monitor or Helm. With --autostart nothing has to run beforehand. The started monitor gets the server's whole environment, so its own database settings (DB_HOST, DB_TLS, DB_TLS_CA_FILE, DB_TLS_CERT_FILE, DB_TLS_KEY_FILE, DB_AUTH_MODE and so on, from the main README) can go in the client's env block. sqlite storage may need better-sqlite3 installed globally (npm install -g better-sqlite3).

Limits

Not covered No direct database access: every tool calls the BetterDB monitor's REST API, so it sees only what the monitor collects and runs no arbitrary commands.
Not covered add_connection and test_connection take host, port, username and password only. TLS, CA bundles and certificate-only (mTLS) sign-in are set on the monitor, not through these tools.
Cost get_anomalies, get_latency_regressions, get_largest_keys and the SLA fields of get_inference_latency are Pro features, free during early access per the README.
Limit COMMANDLOG and per-slot tools need Valkey 8 or newer. Vector tools need valkey-search or RediSearch.
Limit Cache and memory tools work only with BetterDB's semantic-cache, agent-cache or agent-memory libraries in use on the instance.
Not covered memory_recall takes a precomputed embedding vector. The server does not embed text itself.
Limit Each call to the monitor's API times out after 30 seconds (src/index.ts).
Limit With --autostart it gives up if the monitor is not healthy within 15 seconds. --storage takes only sqlite or memory.

It only sees what the BetterDB monitor exposes; it has no direct database access and runs no arbitrary commands. get_anomalies, get_latency_regressions, get_largest_keys and the SLA fields of get_inference_latency are Pro features (free in early access, per the README). COMMANDLOG and per-slot stats need Valkey 8 or newer; vector tools need the Search module (valkey-search or RediSearch); cache and memory tools need BetterDB's own semantic-cache, agent-cache or agent-memory libraries. memory_recall takes a precomputed embedding vector; the server does not embed text. Each API call times out after 30 seconds (src/index.ts). --storage accepts only sqlite or memory.

Who it suits

Suits teams already running BetterDB Monitor, self-hosted or on Cloud, who want an assistant to investigate Valkey or Redis incidents from history that the database itself has discarded: slow commands, hot keys, cluster imbalance, anomalies. Also suits users of BetterDB's semantic or agent caches who want proposed tuning changes reviewed before they are applied. It is a poor fit if you do not want to run the BetterDB monitor at all, or want an assistant to run arbitrary Redis commands or read and write keys directly: this server only goes through the monitor's API.

Tools 60

46 only read, 12 change things, 2 can delete or overwrite. A ? marks an optional parameter.

list_instances: Lists the Valkey/Redis instances the monitor knows, with connection status
select_instance: Chooses the instance later calls work oninstanceId: string
add_connection changes: Registers a new Valkey/Redis connection in BetterDBname: stringhost: stringport?: numberusername?: stringpassword?: stringsetAsDefault?: boolean
test_connection: Checks connection details without saving themname: stringhost: stringport?: numberusername?: stringpassword?: string
remove_connection deletes: Deletes a connection from BetterDBinstanceId: string
set_default_connection changes: Makes a connection the active defaultinstanceId: string
get_health: Summarised health of the selected instance
get_info: INFO output, optionally one sectionsection?: stringinstanceId?: string
get_slowlog: Current slowlog entriescount?: numberinstanceId?: string
get_commandlog: Current COMMANDLOG entries (Valkey 8+)count?: numberinstanceId?: string
get_latency: Latency event list
get_latency_history: History for one latency event typeeventName: stringinstanceId?: string
get_memory: MEMORY DOCTOR and MEMORY STATS output
get_clients: Connected clients and their details
get_hot_keys: Most-accessed keys by LFU or idle timestartTime?: numberendTime?: numberlimit?: numberinstanceId?: string
get_slowlog_patterns: Slowlog grouped into command patterns with counts and averageslimit?: numberinstanceId?: string
get_commandlog_history: Stored COMMANDLOG entries for a time rangestartTime?: numberendTime?: numbercommand?: stringminDuration?: numberlimit?: numberinstanceId?: string
get_commandlog_patterns: COMMANDLOG grouped into patternsstartTime?: numberendTime?: numberlimit?: numberinstanceId?: string
get_anomalies: Detected anomaly events (Pro)limit?: numbermetricType?: stringstartTime?: numberinstanceId?: string
get_client_activity: Client activity over time from stored snapshotsstartTime?: numberendTime?: numberbucketSizeMinutes?: numberinstanceId?: string
get_acl_audit: ACL audit log entriesusername?: stringreason?: stringstartTime?: numberendTime?: numberlimit?: numberinstanceId?: string
get_cluster_nodes: Cluster nodes with roles, health and slot ranges
get_cluster_node_stats: Performance figures per cluster node
get_cluster_slowlog: Slowlog merged across cluster nodeslimit?: numberinstanceId?: string
get_slot_stats: Key counts and CPU per slot (Valkey 8+)orderBy?: enumlimit?: numberinstanceId?: string
start_monitor changes: Launches a local monitor as a background process that outlives the sessionport?: numberstorage?: enum
stop_monitor changes: Stops a monitor started earlier
get_forecast: Projects when memory, ops/sec, CPU or fragmentation hits its ceilingmetricKind: enuminstanceId?: string
get_latency_regressions: Sustained p99 latency rises against baseline (Pro)limit?: numberstartTime?: numberinstanceId?: string
get_largest_keys: Keys using the most memory, from key analytics (Pro)limit?: numberstartTime?: numberendTime?: numberinstanceId?: string
get_vector_indexes: Health of each vector search indexinstanceId?: string
get_inference_latency: FT.SEARCH latency percentiles per index, SLA status on ProwindowMs?: numberinstanceId?: string
ai_list_instances: AI cache and memory components with liveness and latest figuresinstanceId?: string
ai_instance_history: Stored metric history for one AI componentfield: stringhours?: numberinstanceId?: string
list_ai_traces: Recent traces sent in from AI applicationshours?: numberservice?: stringlimit?: number
get_ai_trace: All spans of one trace with timingstraceId: string
correlate_ai_trace: Matches a trace's cache and memory spans to live Valkey statetraceId: stringinstanceId?: string
cache_list: Caches registered on the instance with hit rate and op countsinstanceId?: string
cache_health: Health details for one cachecache_name: stringinstanceId?: string
cache_threshold_recommendation: Suggested similarity threshold for a semantic cachecache_name: stringcategory?: stringminSamples?: numberinstanceId?: string
cache_tool_effectiveness: Per-tool hit rate, savings and TTL advice for an agent cachecache_name: stringinstanceId?: string
cache_similarity_distribution: Histogram of recent similarity scores for a semantic cachecache_name: stringcategory?: stringwindow_hours?: numberinstanceId?: string
cache_recent_changes: Recent proposals for one cachecache_name: stringlimit?: numberinstanceId?: string
cache_propose_threshold_adjust changes: Files a proposal to change a similarity thresholdcache_name: stringnew_threshold: numbercategory?: stringreasoning: stringinstanceId?: string
cache_propose_tool_ttl_adjust changes: Files a proposal to change a tool's TTLcache_name: stringtool_name: stringnew_ttl_seconds: numberreasoning: stringinstanceId?: string
cache_propose_invalidate changes: Files a proposal to invalidate cache entriescache_name: stringfilter_kind: enumfilter_expression?: stringfilter_value?: stringestimated_affected: numberreasoning: stringinstanceId?: string
cache_list_pending_proposals: Proposals awaiting a decisioncache_name?: stringlimit?: numberinstanceId?: string
cache_get_proposal: One proposal with its audit trailproposal_id: string
cache_approve_proposal changes: Approves a proposal and applies it to Valkeyproposal_id: stringactor?: string
cache_reject_proposal changes: Rejects a proposal, with an optional reasonproposal_id: stringreason?: stringactor?: string
cache_edit_and_approve_proposal changes: Changes a threshold or TTL proposal's value and applies itproposal_id: stringnew_threshold?: numbernew_ttl_seconds?: numberactor?: string
memory_stores: Agent-memory stores found on the instance
memory_list: Memories in a store, newest first, filterable by scope and tagmemory_name: stringthreadId?: stringagentId?: stringnamespace?: stringtags?: arraylimit?: numberoffset?: numberinstanceId?: string
memory_get: One memory by IDmemory_name: stringid: stringinstanceId?: string
memory_stats: Item and eviction counts and config of a memory storememory_name: stringinstanceId?: string
memory_recall: Finds memories by a supplied embedding vectormemory_name: stringvector: arrayk?: numberthreshold?: numbertags?: arraythreadId?: stringagentId?: stringnamespace?: stringinstanceId?: string
memory_forget changes: Files a proposal to delete memories by ID, scope or tagsmemory_name: stringid?: stringthreadId?: stringagentId?: stringnamespace?: stringtags?: arrayreasoning: stringinstanceId?: string
memory_list_pending_forgets: Delete proposals awaiting approvalmemory_name?: stringlimit?: numberinstanceId?: string
memory_approve_forget deletes: Approves a delete proposal and removes the memoriesproposal_id: stringactor?: stringinstanceId?: string
memory_reject_forget changes: Rejects a delete proposalproposal_id: stringreason?: stringactor?: stringinstanceId?: string

Popularity

43 popularity score, on our list (the bar is 36)

★ 1.3k ↓ 75/wk

GitHub stars, whole repository	1,312	15 points
Stars in 30 days	+8	4.8 points
npm downloads, last week	75	15 points
Official server	Yes	8 points

It lives in the folder packages/mcp of betterdb-inc/monitor, which holds more than this server. Its stars are the whole repository's, so in the score they never count for more than its own downloads.

Measured 2026-11-07. How the score works.

Where it is published

Code	betterdb-inc/monitor/packages/mcp ★ 1.3k, the whole repository
npm	@betterdb/mcp 1.5.0, 2026-08-31

What it can reach

Runs	On your machine
Reaches	The internet, A database, Runs shell commands
Changes things	Yes: it can write, send, create or delete
Sign-in	An API key or token
Licence	MIT (npm)
Usage data	Sends it unless you turn it off (`BETTERDB_TELEMETRY=false`) Tool name, success, duration and any error message per call, batched every 30 seconds and posted to the monitor's /mcp/telemetry endpoint at BETTERDB_URL (src/telemetry.ts)
Last code change	2026-11-06

Runs on your machine over stdio and makes HTTP calls to the BetterDB monitor at BETTERDB_URL, sending BETTERDB_TOKEN as a Bearer header when set. add_connection and remove_connection edit the monitor's stored connections, including database passwords. cache_approve_proposal, cache_edit_and_approve_proposal and memory_approve_forget apply changes or deletions to live Valkey data, while the propose tools only file proposals. Unless BETTERDB_TELEMETRY is false it reports tool names, timings and errors to that same monitor. With --autostart or start_monitor it spawns npx --yes @betterdb/monitor, passing on its full environment, and writes ~/.betterdb/monitor.pid. With --persist the monitor keeps running after the session until stop_monitor or --stop.

Over time

GitHub stars, whole repository 1.3k +8 (+1 %)

1.3k1.3k

8 Oct 20267 Nov 2026

npm downloads a week 75 −93 (−55 %)

3633

23 Mar 20267 Nov 2026

Before the site measured it, read from npm's daily download counts.

Popularity score 43 −9 (−18 %)

5227

8 Oct 20267 Nov 2026

What is rising across the directory

Sources

Page checked 2026-11-05.

Badge for your README

Shows this server's rank in its category, and stars for a server that runs on your machine. It updates every hour.

MarkdownREADME.md

[![BetterDB MCP server on EveryMCPServer](https://everymcpserver.com/badge/monitor.svg)](https://everymcpserver.com/servers/monitor)

Connect your client

Add in one click

Or add it by command or config

Settings 4

What it does

What it needs

Limits

Who it suits

Tools 60

Similar servers

Better Stack MCP Server

InfluxDB 3 MCP Server

Render MCP Server

MCPQL

Confluent MCP Server

Azure MCP Server