Your Postgres estate monitored, diagnosed, and under control.
Introducing the pgEdge AI DBA Workbench
AI-powered monitoring and management purpose-built for PostgreSQL. Continuous metrics collection, three-tier anomaly detection, and an agentic AI assistant that works through problems with you.
Triple your Postgres productivity with an always-on co-pilot.
Most teams want to do more with PostgreSQL. The bottleneck is the shortage of skilled Postgres people to run it. The pgEdge AI DBA Workbench is a co-pilot and force multiplier for the DBAs you already have. It watches every instance and catches anomalies before they become outages. It walks through diagnosis and a resolution step by step, so a senior DBA can cover more ground and a junior DBA can handle cases they couldn’t before.
Runs on ANY PostgreSQL 14+ instance, anywhere — including Amazon RDS, Supabase, and pgEdge Cloud. No migration. No rip-and-replace.
Four components. One self-hosted platform.
Your data never leaves your infrastructure.
Collector: Continuous metrics
34 configurable probes across every monitored instance. Replication lag, vacuum, query stats, WAL throughput, connections, system resources, security posture, and more. Connects remotely over standard PostgreSQL wire protocol. No agent installed on the database host.
Server: MCP endpoint & REST API
HTTP/HTTPS with JSON-RPC 2.0. SQLite-based auth with role-based access control. LLM proxy supports Antropic, OpenAI, Gemini, Ollama, and other local model runners with an OpenAI compatible API such as LM Studio and llama.cpp. Manages conversation history.
Alerter: Three-tier anomaly detection
Thresholds and statistical baselines, embedding similarity, and LLM reasoning for the ambiguous cases. Alerts clear automatically when conditions normalise. Blackout scheduling with cron recurrence for maintenance windows.
Client: Browser-based UI
Drill-through dashboards from estate down to individual index stats. Cluster navigator with colour-coded status across nodes. Full admin panel for users, tokens, connections, and alert rules.
Meet Ellie
An AI agent that diagnoses, not just describes.
Ellie has 21 MCP tools and can run EXPLAIN ANALYZE, inspect schemas, query historical metrics, and walk through multi-step diagnostic workflows. She has a comprehensive knowledgebase of PostgreSQL and related projects, as well as pgEdge projects ensuring she has all the information she needs to fully understand what she’s working with, regardless of model training. When she finds an issue, she provides the SQL to fix it. You review it. You decide. If it looks good, simply press play to have Ellie execute the SQL, or copy and paste to psql or your favorite tool.
Ellie remembers. Pin context like "our busiest period is 2-4pm EST" and it carries into every conversation.
Run EXPLAIN ANALYZE and interpret the output
Query historical metrics against current state
Inspect schemas, indexes, and table statistics
Provide specific SQL to resolve identified issues
Persistent memory with semantic recall
Three tiers because one is not enough.
Fixed thresholds catch the obvious. The Workbench catches the rest.Statistical baselines
Establishes normal behavior per metric per instance. Catches deviations without manual threshold configuration. Adapts as your workload changes.
Embedding similarity
Matches metric patterns against a library of known anomaly signatures using vector embeddings. Catches patterns that statistical rules miss, such as periodic locking storms.
LLM classification
For ambiguous cases, an LLM evaluates with full context: metric history, cluster state, and recent events. Reasons about what is happening, not just that something changed.
What's included
Everything your Postgres estate needs. One install.
Monitoring & Dashboards
34 built-in probes: replication, queries, connections, vacuum, WAL
Drill-through from estate to cluster to server to database to individual indexes
Cluster navigator with status indicators across nodes
AI-generated summaries at estate, cluster, and server level
MCP Server
HTTP/HTTPS with JSON-RPC 2.0
Standards-compliant MCP, works with Claude Code, Cursor, VS Code + Copilot, Codex, or any HTTP MCP client
Role-based access control and API token support
Same tools and resources Ellie uses, exposed to any MCP client
Alerting
Hierarchical thresholds: tighter bounds for production, relaxed for staging
Automatic alert clearing when conditions normalize
Blackout scheduling with cron recurrence
Slack, email, Mattermost, and webhook notifications
Security
AES-256-GCM encryption for stored credentials
Role-based access control with group and privilege management
Read-only modes for safe exploration
Air-gapped deployment: no data leaves your network
AI Assistant (Ellie)
21 MCP tools: metrics, schema inspection, EXPLAIN analysis
Multi-step workflows up to 50 tool iterations
SQL recommendations reviewed and applied by the administrator
Persistent memory with semantic recall
LLM Flexibility
Anthropic Claude, OpenAI, Google Gemini for cloud inference
Ollama natively, plus any OpenAI-compatible endpoint (LM Studio, llama.cpp, EXO) for fully local, air-gapped inference
Switch providers without changing your workflow
AI Optional
Run without AI features if you prefer
Gracefully disables AI features to operate as a “traditional” management system
Your AI dev tools, connected to your monitoring data.
Any HTTP MCP-compatible client gets the same access Ellie has: live metrics, schema introspection, alert history, query execution, and EXPLAIN ANALYZE. Investigate a slow query in Cursor and immediately see the cluster context behind it. Works the same whether the database is community Postgres, pgEdge, RDS, or Supabase.
Compatible clients: Claude Code | Cursor | VS Code + Copilot | Codex | Any MCP client
Runs on your infrastructure. Your data stays with you.
Any PostgreSQL 14+ instance, anywhere! No agent installation required.
Any PostgreSQL 14+ instance
Community Postgres, pgEdge Enterprise Postgres, pgEdge Cloud, and any Postgres DBaaS including Amazon RDS, Azure Database for PostgreSQL, Google Cloud SQL, and Supabase. No migration required.
Air-gapped and on-premises
Local model runners (Ollama, LM Studio, llama.cpp, EXO via OpenAI-compatible endpoints) for fully air-gapped deployments. Meets government and regulated-environment requirements.
Agentless collection
Remote collection over standard PostgreSQL connections. Add instances to monitoring without touching the host server.
Open source, PostgreSQL License
Free on GitHub. pgEdge Enterprise Postgres subscribers receive support at no extra cost.
How it compares
The only Postgres monitor built for AI-assisted operations.| Capability | pgEdge AI DBA Workbench | EDB PEM | pganalyze | Datadog DBM | Percona PMM | pgwatch |
|---|---|---|---|---|---|---|
| AI / LLM assistant | Agentic, 21 tools, multi-step | None | Index advisor only | ML anomaly detection | Built-in advisors | None |
| MCP protocol | Native | No | Preview | No | No | No |
| Distributed PG awareness | Native, multi-master detection and monitoring | Patroni only | Generic | Generic | Generic | Generic |
| Anomaly detection | 3-Tier (Stats + embeddings + LLM) | Threshold-based | Threshold-based | ML-based | ML-based (v3+) | Threshold |
| Self-hosted | Yes | Yes (subscription) | Docker | SaaS platform | Yes | Yes |
| Local / Air-gapped LLM | Yes (Ollama, LM Studio, llama.cpp, EXO) | N/A | N/A | N/A | N/A | N/A |
| Persistent AI memory | Yes | No | No | No | No | No |
| Agent required | No | Yes (per-server) | Yes | Yes | Yes | No |
| Pricing | Free (OSS) | EDB subscription | $149-$399+/mo | From $70/host/mo | Free (OSS) | Free (OSS) |
Ready to give your Postgres estate a co-pilot?
Fully open source. Self-hosted. PostgreSQL-licensed. Available now on GitHub.Dive deeper into pgEdge
Postgres Live: Monthly Series Scaling PostgreSQL: Vertical, Horizontal, and High Availability Strategies
Introducing The pgEdge Postgres MCP Server - And How to Connect it to Claude Code and Cursor
Get started today.
Experience the magic of pgEdge for non-distributed and distributed Postgres deployments.