Petabyte-scale data doesn’t have to mean petabyte-scale complexity. But for most teams, that’s exactly how it feels. Analysts write increasingly complex queries. Engineers jump between tools. Business stakeholders wait days for answers that should take seconds.
The Model Context Protocol (MCP) changes that by giving teams a natural language interface to their data. Instead of writing SQL or navigating dashboards, anyone can ask a question and get an answer—directly—regardless of data volume.
That’s exactly what the Hydrolix MCP Server enables.
Since its initial release in 2025, teams have been using Hydrolix MCP to slash query times, eliminate tool-switching, and put data access directly into the hands of analysts and business users. Today, we’re expanding how teams can deploy it.
In addition to running the MCP server locally on a developer’s machine, you can now deploy it server-side, directly within your Hydrolix cluster.
With server-side deployment, teams can connect any MCP-compatible AI assistant— Claude, Gemini, or whichever tool they prefer—through a simple URL. The experience is consistent across the organization: database discovery, schema exploration, and natural language querying against full-fidelity log data.
If local deployment already works well for your team, nothing changes. But if you’re looking for easier organization-wide access or a deployment model that aligns with enterprise security and policy requirements, server-side MCP was built for you.
With Hydrolix Remote MCP, anyone in your organization can securely connect MCP-compatible AI tools to complete, unsampled log data—no local installs required.
We’ve also added performance-focused enhancements. The MCP server now supports automatic switching between raw and summary tables, enabling faster queries across petabyte-scale datasets without sacrificing precision. Additional updates include new tools and optimized token usage to improve query efficiency.
And to see what this looks like in the real world, consider an incident that was resolved in less time than it takes to bake a pizza.
Hydrolix MCP Leads to Faster Incident Resolution
A major enterprise was experiencing service degradation across multiple services behind a load balancer. The root cause? A capacity issue causing devices to repeatedly attempt connections, cascading into broader impact for end users.
Because the MCP server was already running server-side in their Hydrolix cluster, there was nothing to install or configure. The engineer on call simply connected their AI assistant and asked:
“I’m seeing an increase in 5XX errors and need help with an investigation. Look into this and determine if there is a shared pattern among the request paths where these elevated errors are occurring.”
The AI queried their full-fidelity log data and immediately surfaced the pattern without SQL, context-switching, or digging through dashboards.
Just a question and an answer. Within minutes, the engineer had pinpointed the correct domain and identified the fix. The total time from joining the call to mitigation was less than 10 minutes.
The following screenshots show how the engineer used natural language to investigate.
In the previous screenshot, the engineer investigates patterns in affected request paths and the MCP server queries the Hydrolix tables.
In the next screenshot, the AI tool returns key findings from the queries that the MCP server ran and analyzes the results to isolate the troublesome endpoint.
That exchange says it all. When your AI assistant has direct access to complete, unsampled log data that’s always hot and queryable, investigations that used to take hours can be resolved before the pizza arrives.
Why Full-Fidelity Data Matters
Most observability platforms make a quiet trade-off: they downsample, aggregate, or shuffle older data off to cold storage to manage costs. That works fine until you need to find something subtle—a pattern buried in specific request paths, an anomaly hiding in individual transactions. At that point, sampled data has gaps, and aggregated dashboards don’t have the granularity. You’re left guessing.
Hydrolix doesn’t make that trade-off. Every piece of data stays raw, granular, and immediately queryable—no cold storage to thaw, no gaps from sampling. In the incident above, that mattered. The pattern wasn’t obvious. It required full-fidelity data to surface. Traditional platforms that rely heavily on sampled data wouldn’t have revealed the issue.
Full-fidelity data is only useful if you can get complete insights from it, but at petabyte-scale, that has traditionally required specialized skills and significant time. MCP removes that barrier. Any engineer, analyst, or on-call responder can query complete, unsampled logs through an AI assistant and get a direct answer, without writing SQL or waiting for someone who can. The result: deeper insights, available to everyone on your team.
Next Steps
Interested in trying out Hydrolix? Contact us about a managed trial or demo.
