Postgres Scout MCP
Scout your PostgreSQL databases with AI - A production-ready Model Context Protocol server with built-in safety features, monitoring, and data quality tools.
接入你的智能体
發布者自己的指令。不是 PlugMCP 端點。
用我們的金鑰與日誌託管一份副本。
npx -y [email protected]Claude Code 沒有本機指令的安裝連結。複製下方指令並在終端機中執行。
claude mcp add io.github.bluwork-postgres-scout-mcp -- npx -y postgres-scout-mcp@1.0.3在終端機中執行。
{ "mcpServers": { "io.github.bluwork-postgres-scout-mcp": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}claude_desktop_config.json
遠端 URL 還是本機指令? 兩種安裝方式的差異
You ask:
"How healthy is my production database? Any urgent issues?"
Postgres Scout returns:
Component Breakdown
Issues Found
orders has 34% bloat (2.1 GB wasted). VACUUM FULL recommended.sessions consuming 890 MB.analytics_events is 71% (target: >90%).Recommendations
VACUUM FULL orders during maintenance windowidx_sessions_legacy, idx_sessions_old_token, idx_sessions_tempanalytics_events to shared_buffers or partitioning by dateThat's getHealthScore — one of 38 tools covering exploration, diagnostics, optimization, monitoring, data quality, and safe writes.
claude mcp add postgres-scout -- npx -y postgres-scout-mcp postgresql://localhost:5432/mydb
Then ask: "Show me the largest tables and whether they have any bloat issues."
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
{
"mcpServers": {
"postgres-scout": {
"command": "npx",
"args": ["-y", "postgres-scout-mcp", "postgresql://localhost:5432/mydb"],
"type": "stdio"
}
}
}
Add to your MCP settings:
{
"postgres-scout": {
"command": "npx",
"args": ["-y", "postgres-scout-mcp", "postgresql://localhost:5432/mydb"]
}
}
The server runs in read-only mode by default. For write operations, run a separate instance:
{
"mcpServers": {
"postgres-scout-readonly": {
"command": "npx",
"args": ["-y", "postgres-scout-mcp", "--read-only", "postgresql://localhost:5432/production"],
"type": "stdio"
},
"postgres-scout-readwrite": {
"command": "npx",
"args": ["-y", "postgres-scout-mcp", "--read-write", "postgresql://localhost:5432/development"],
"type": "stdio"
}
}
}
listDatabases — databases the user has access togetDatabaseStats — size, cache hit ratio, connection infolistSchemas — all schemas in the current databaselistTables — tables with size and row statisticsdescribeTable — columns, constraints, indexes, and moreexecuteQuery — run SELECT queries (or writes in read-write mode)explainQuery — EXPLAIN plans for performance analysisoptimizeQuery — optimization recommendations for a specific querygetHealthScore — overall health score with component breakdowndetectAnomalies — anomalies in performance, connections, and dataanalyzeTableBloat — bloat analysis for VACUUM planninggetSlowQueries — slow query analysis (requires pg_stat_statements)suggestVacuum — VACUUM recommendations based on dead tuples and bloatsuggestIndexes — missing index recommendations from query patternssuggestPartitioning — partitioning strategies for large tablesgetIndexUsage — identify unused or underused indexesgetCurrentActivity — active queries and connectionsanalyzeLocks — lock contention and blocking queriesgetLiveMetrics — real-time metrics over a time windowgetHottestTables — tables with highest activitygetTableMetrics — comprehensive per-table I/O and scan statsfindDuplicates — duplicate rows by column combinationfindMissingValues — NULL analysis across columnsfindOrphans — orphaned records with invalid foreign keyscheckConstraintViolations — test constraints before adding themanalyzeTypeConsistency — type inconsistencies in text columnsexploreRelationships — multi-hop foreign key traversalanalyzeForeignKeys — foreign key health and performancefindRecent — rows within a time windowanalyzeTimeSeries — window functions and anomaly detectiondetectSeasonality — seasonal pattern detectionexportTable — CSV, JSON, JSONL, or SQLgenerateInsertStatements — INSERT statements for migrationpreviewUpdate / previewDelete — see what would change before committingsafeUpdate — UPDATE with dry-run, row limits, empty WHERE protectionsafeDelete — DELETE with dry-run, row limits, empty WHERE protectionsafeInsert — INSERT with validation, batching, ON CONFLICT support"What are the largest tables and do they have bloat?"
listTables({ schema: "public" })
analyzeTableBloat({ schema: "public", minSizeMb: 100 })
"Find duplicate emails in the users table."
findDuplicates({ table: "users", columns: ["email"] })
"Which queries are slowest and how can I speed them up?"
getSlowQueries({ minDurationMs: 100, limit: 10 })
suggestIndexes({ schema: "public" })
"Show me what's happening on the database right now."
getCurrentActivity()
getLiveMetrics({ metrics: ["queries", "connections", "cache"], duration: 30000, interval: 1000 })
getHottestTables({ limit: 5, orderBy: "seq_scan" })
"Find orphaned orders that reference deleted customers."
findOrphans({ table: "orders", foreignKey: "customer_id", referenceTable: "customers", referenceColumn: "id" })
CLI flags: --read-only (default), --read-write, --mode <mode>
File logging is disabled by default. Set ENABLE_LOGGING=true to enable. Two log files are created in LOG_DIR:
Connection strings are automatically redacted in all output.
git clone https://github.com/bluwork/postgres-scout-mcp.git
cd postgres-scout-mcp
pnpm install
pnpm build
pnpm test
Apache-2.0
Scout your PostgreSQL databases with AI - safety features, monitoring, and data quality
{ "mcpServers": { "io.github.bluwork-postgres-scout-mcp": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式,而不是 PlugMCP 代理伺服器。
這筆記錄未列出必要的環境變數。
Postgres Scout 透過 stdio 在本機執行。
上次檢查時間:2026-11-09。 Scout your PostgreSQL databases with AI - safety features, monitoring, and data quality