Observe AI SRE

Note

This is a private preview feature. Please contact your Observe representative for more information or open Docs & Support → Contact Support in the product and let us know.

Observe AI SRE is your AI-powered observability assistant. It helps you investigate incidents, debug slow or error-prone services, and summarize observability insights across logs, metrics, and traces all through natural language.

AI SRE leverage Claude and OpenAI models securely. It leverages the same Observe MCP Server used by agentic tools like Augment, Cursor, Claude Code, Claude Desktop, OpenAI Codex CLI, and Windsurf, so that AI SRE can reason directly over your telemetry data.

How it works

  1. You ask a question in the o11y chat pane, such as
    “Why did latency spike last night?” or “Show the slowest services in my cluster.”

  2. AI SRE interprets your intent using Claude Sonnet 4.5, and calls the Observe MCP Server.

  3. The MCP Server accesses your Knowledge Graph Tool and Query Generation Tool to find relevant datasets (logs, metrics, traces, resources, correlations).

  4. Observe executes the query and returns summarized insights directly in the chat.

What data leaves Observe?

Observe AI SRE architecture

AI SRE uses the same privacy-preserving Zero Data Retention (ZDR) mechanisms as the Observe MCP Server.
All requests are processed in-memory by the model providers and never stored or used for training.

Data sent to OpenAI

Event

What we send to OpenAI

Purpose

Knowledge graph build / refresh
(runs periodically)

• Dataset schema (names & types)
• ≤ 10 sample rows per dataset
• Column‑level stats
• Metric names & discovered tag keys/values

Allows the agent to reason over your dataset catalogue.

Data sent to Claude

Event

What we send to Claude

Purpose

Runtime query
(each time an agent executes an query)

• A short dataset summary document for the candidate tables (no raw rows)

Helps the model choose the correct dataset(s) to answer the user’s question.

User query execution
(each time a user asks AI SRE a question)

• User’s natural-language prompt
• System prompt instructions (not retained or logged)
• Structured query results

Allows Claude to generate summaries, diagnostic explanations, and root-cause analyses based on Observe query results.

OpenAI and Claude don’t train their models on your data by default, but we’d like to be transparent about the data that is sent to OpenAI and Claude when you use thr Observe AI SRE. In addition, all requests use Zero Data Retention (ZDR) for both OpenAI and for Claude — data is processed in-memory by OpenAI and Claude and never stored or logged.

Prerequisites

Minimum RBAC permission

AI SRE cannot see or query any dataset that the user’s Observe role does not already grant access to.

Try these questions

Common questions in Observe AI SRE 
Why did error rate increase after deployment?
Show top 5 services by latency this morning.
Which Lambda functions are slow or error-prone?
What resources had CPU bottlenecks?
Summarize service health in the last 24 hours.