Self-Hosted Team AI SRE with Episodic Memory

OpenSRE — the memory-first AI SRE — is a self-hosted platform that investigates production incidents, remembers what it learns, and maps service topology. This post is for teams who want that capability inside their own network, with a real console and chat entry points — not only a laptop CLI.

What “memory-first” means

Episodic memory stores structured outcomes from past investigations. When a similar alert fires, OpenSRE can recall prior root causes and approaches instead of starting from zero. That is the difference between a one-shot agent and an SRE teammate that accumulates institutional knowledge.

Knowledge graph for blast radius

A Neo4j knowledge graph gives investigations service dependencies and blast radius context. Symptom chasing without topology is how on-call burns hours. Topology-aware investigation is how you stop guessing.

Team surfaces, not only a single developer shell

Teams need a web console for Investigations, hierarchical org→team configuration, and chat in Slack and Microsoft Teams. OpenSRE at opensre.in ships those as first-class surfaces on a stack you run with Docker Compose or Kubernetes.

Same name, different product

If you searched for “OpenSRE” and found more than one project: this one lives at opensre.in. It is not affiliated with Tracer / opensre.com. Feature-by-feature: OpenSRE vs Tracer OpenSRE.

Get started

See the quick start or the GitHub repository.