HolmesGPT: A CNCF Sandbox Agent for Root-Cause Investigations and Operator Mode
SRE Agent - CNCF Sandbox Project
At a glance
- What is it?
- HolmesGPT is an open-source Python agent that investigates production incidents by querying live observability data, and its Operator mode runs 24/7 health checks. This review covers its architecture, setup, limitations, and fit for SRE teams.
- Who is it for?
- HolmesGPT can help an operations team investigate incidents by querying configured data sources and producing an analysis for review. The operator runs in Kubernetes, while the core CLI can investigate other infrastructure.
- Can I use it commercially?
- Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository received new commits within the last day.
- What is it written in?
- Mainly Python, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- HolmesGPT/holmesgpt3,902 ★ · +640
| Repository | 2026-09-10 | 2026-11-15 |
|---|---|---|
| HolmesGPT/holmesgpt | 3.3k | 3.9k |
640 Stars in the last 66 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
The Problem: Incident Response Still Starts with a Human
The problem is not just detection. It is also the investigation itself. An SRE often has to jump between Prometheus, Grafana, logs, and cloud dashboards to piece together what happened. HolmesGPT aims to automate that by using an agentic loop. The README describes it as an "agentic loop to query live observability data from multiple sources and identify root causes." That means it does not just fetch a single metric; it iterates, asks follow-up questions, and correlates information. The design assumes that the bottleneck is not data availability but the cognitive work of connecting signals. For teams with large fleets, that cognitive work does not scale. An agent that can do it in parallel, or in the background, is the intended answer.
How the Agentic Loop Works with Built-In Toolsets
The project documents integrations that fetch alerts or tickets and write an analysis back to the source or Slack. Availability differs by integration: its README lists Slack and Teams through Robusta, while several other integrations are listed for the HolmesGPT CLI. Verify the exact read and write operations in current documentation; a write-back capability does not establish that every integration can acknowledge or close an alert. Grant only the permissions needed and audit actions.
Operator Mode: Scheduled Health Checks and Deployment Verification
Operator mode runs in Kubernetes and supports scheduled health checks and deployment verification against connected data sources. The README describes an optional GitHub integration that can open pull requests. A proposed change is not a verified fix: review permissions, generated code and tests before merging, and check current integration documentation for the available guardrails.
Getting It Running: Installation and Configuration
Start with the project's installation and provider guides, then configure only the toolsets needed for a pilot. The core CLI can be used without Kubernetes; operator mode runs in Kubernetes. Check the documentation for the exact version's configuration keys, authentication and data handling, and test the integration with a non-critical service before granting wider access.
Memory-Safe Execution and Data Scaling: The Real Engineering
The README describes filtering, JSON traversal and output transformers for reducing what enters the model context, plus memory limits, disk streaming and output budgets for large tool results. These mechanisms can reduce resource pressure, but the project does not establish that all workloads avoid out-of-memory failures or that filtered output retains the decisive evidence. Test limits, storage and retrieval quality against representative incidents.
Limitations and Failure Modes: When It Is the Wrong Tool
A third limitation is the toolset coverage. While the list is long, it is not exhaustive. If your stack uses a niche observability tool that is not listed, you will need to build a custom toolset or use the REST API toolset. That requires development effort. The README does not quantify the effort. For a small team, that could be a significant barrier. The project is also a CNCF sandbox project, which means it is early-stage. APIs and features can change between releases. The recent releases (0.41.0, 0.40.0) show active development, but that also means you need to track upgrades for breaking changes.
Alternatives and Comparisons: What Else Is Out There
Rule-based automation and scheduled checks can also correlate signals and identify known failure patterns when operators encode those rules. HolmesGPT adds model-assisted investigation across configured data sources, with variable model cost and potentially inconsistent results. Compare both approaches using representative incidents, measurable accuracy, access controls and human review requirements.
Editorial conclusion
HolmesGPT can help an operations team investigate incidents by querying configured data sources and producing an analysis for review. The operator runs in Kubernetes, while the core CLI can investigate other infrastructure. Start with limited permissions and a non-critical pilot, verify data sent to model providers, and measure investigation quality and cost. Check integration-specific permissions and keep human review for proposed fixes and actions.
Frequently asked questions
What is HolmesGPT?
An open source AI agent for investigating production incidents and finding root causes, using an agentic loop to query live observability data across multiple sources. It is a CNCF sandbox project, originally created by Robusta.Dev with major contributions from Microsoft.
what is holmesgpt
HolmesGPT is an open-source incident investigation agent that can query configured toolsets across Kubernetes and other infrastructure. The core CLI does not require Kubernetes; check the documentation for the supported operations and configuration of each data source or REST integration.
How does HolmesGPT avoid running out of memory or context?
The project describes server-side filtering, output transformations, per-tool memory limits, disk streaming and output budgets to manage large results. These controls reduce resource and context pressure but do not guarantee that every workload avoids memory exhaustion or preserves every relevant signal.
What is HolmesGPT operator mode?
Operator mode runs health checks in Kubernetes and can report findings. With a configured GitHub integration it can propose pull requests; detection and fixes are not guaranteed, and permissions and review should be configured for the deployment.
Which model providers does HolmesGPT support?
OpenAI, Anthropic, Azure, Bedrock, Gemini and more. They are reached through one routing library whose version is pinned to an exact release, with comments recording which advisory that pin addresses.
Official sources
Was this useful?