November 12, 2026: PostgreSQL 18.7, 17.12, 16.16, 15.20, 14.25, and 19 RC 2 Released!

Datasentinel 2026.10 released with AI Connector

Posted on 2026-11-15 by Datasentinel
Related Proprietary

Datasentinel 2026.10, the new release of our PostgreSQL monitoring and troubleshooting platform, is now available.

It introduces AI Connector, which connects Datasentinel metrics to the AI assistant your team already uses.

Ask a question in plain language, such as "why was the orders database slow at 9:00 this morning?". The assistant analyzes the collected data, explains what it finds and turns the evidence into concrete recommendations.

AI Connector

AI Connector works with the main AI assistants, including Claude, ChatGPT, Gemini, Microsoft Copilot, GitHub Copilot, and you choose the AI model behind them. It combines an MCP server, which gives the assistant access to Datasentinel metrics, with a Skill, which guides its analysis. It is an add-on, enabled by a licence option.

What you can ask for:

  • Past performance. Find out what happened on one instance or a group of instances, from slow queries to locks.
  • Live troubleshooting. See what an instance is doing right now: running sessions, who is blocking whom, and how a query would be executed.
  • Comparisons. Compare two periods, instances or applications, for example before and after a deployment.
  • Recommendations. Get concrete advice on queries, indexes, settings and maintenance, based on your own metrics.
  • Several ways to use it. The common way is to chat with your assistant, from a desktop app, a command line or an IDE. You can also generate dashboards on demand, for example as Claude artifacts, schedule health checks and digests, produce HTML reports, or call the MCP server from your own scripts and agents.

Answers adapt to the reader. The assistant is technical with experts and uses plain language with everyone else, in the user's own language.

Examples

Questions are asked in plain language, as you would ask a colleague:

  • "What happened on prod-eu-1 between 08:00 and 10:00 this morning?"
  • "prod-eu-1 feels slow, find out what's wrong."
  • "Compare prod-eu-1 today 08:00-10:00 vs yesterday same window: did it regress?"
  • "Any missing or unused indexes on prod-eu-1? Is autovacuum keeping up?"
  • "Every day, summarize yesterday's workload for my prod-eu-1 instance and flag any regressions."

The assistant answers with findings ordered by impact, each with the evidence behind it and a concrete next step. Follow-up questions keep the context, so an investigation can go from a fleet overview down to one query plan in a few exchanges.

More worked examples are in What you can analyze.

Security model

Opening monitoring data to an AI assistant does not open the databases:

  • Secure sign-in. Users sign in on the Datasentinel login page (OAuth 2.1). The assistant gets a token, never a password or a database credential.
  • Same rights as the web interface. Users see only the instances their role allows. Live analysis requires a dedicated privilege.
  • No direct database access. The assistant never connects to a database. Live checks are read-only and go through the Datasentinel agent. Execution plans are estimated without running the query.
  • What the AI sees. Metrics, SQL text, execution plans and object names reach the assistant and the LLM you choose.
  • Scripts and audit. Scripts and scheduled jobs use access tokens that can be revoked at any time. Sign-ins and token events are logged.

Other changes in 2026.10

  • PDF reports. New, more compact layout. The Workload report shows the load profile and top wait events. Reports now include blocking scenarios, wait events, collected parameters, and more table and index details. They use less memory and generate faster on large fleets.
  • Internal function. The new optional datasentinel_analytics function gives read-only access to planner statistics and execution plans. It requires Agent v4.0.0 or later.
  • Documentation MCP server. Assistants can answer product questions directly from the Datasentinel manual.
  • Inactive instances. Instances with no data left in the metrics database can now be purged automatically.
  • Agentless monitoring. Query monitoring filters can be set once for all agentless instances.
  • Security. Third-party dependencies updated to address known vulnerabilities.

Try it

  • Public sandbox. Try AI Connector without installing a platform, with a shared, read-only account. See Test with the sandbox platform to get started.
  • Your own instances. Request a trial at contact@datasentinel.io.

Links

About Datasentinel

Datasentinel is a commercial platform for monitoring and troubleshooting PostgreSQL performance.

The Datasentinel team also maintains pg_datasentinel, an open source observability extension for PostgreSQL.

For more information, please contact us.