Holmesgpt
SRE Agent - CNCF Sandbox Project
Last verified:
What is Holmesgpt?
HolmesGPT is an open-source AI troubleshooting agent built for Kubernetes and cloud-native environments. It is a Cloud Native Computing Foundation (CNCF) Sandbox project accepted in October 2025. The tool combines observability telemetry (logs, metrics, traces), LLM reasoning, and structured runbooks to accelerate root cause analysis and suggest actionable remediations for DevOps and SRE teams.
Unlike static dashboards or passive chatbots, HolmesGPT is agentic: it actively decides what data to fetch, runs targeted queries across Kubernetes, Prometheus, Datadog, Elasticsearch, AWS, GCP, databases, and more, then iteratively refines hypotheses while staying within your environment. It automatically gathers context, identifies root causes, and provides specific steps to fix problems without manual troubleshooting.
Key features include terminal CLI access, a web interface/TUI, K9s plugin integration, Operator Mode that runs 24/7 in the background to spot problems before customers notice, Slack messaging with fixes, GitHub integration for auto-opening PRs, automatic correlation of alerts, investigation capabilities across PagerDuty and OpsGenie, and benchmarks comparing LLM performance across 150+ test scenarios. It supports multiple AI providers including OpenAI and Anthropic models.
HolmesGPT is designed for DevOps engineers, site reliability engineers (SREs), on-call engineers, and anyone debugging production incidents in cloud-native environments. It helps teams solve alerts faster and reduces the time spent piecing together clues during production debugging.
The tool is open-source and free to use. Users only pay for the AI provider API keys they choose (OpenAI, Anthropic, or any OpenAI-compatible LLM). There are no paid tiers for HolmesGPT itself since it's an open-source CNCF Sandbox project.
Holmesgpt pricing
Pricing model: Freemium
HolmesGPT is open-source and completely free to use. It is a CNCF Sandbox project with no paid tiers. Users only need to pay for their chosen AI provider API keys (OpenAI, Anthropic, or any OpenAI-compatible LLM). OpenAI is the default provider. New OpenAI accounts may receive $5 in credits. There are no HolmesGPT subscription fees or enterprise plans since the tool itself is free and self-hosted.
Holmesgpt pros
- Open-source and free to use (CNCF Sandbox project)
- Agentic AI that actively decides what data to fetch
- Supports 150+ test scenarios for LLM benchmarking
- Works with multiple AI providers (OpenAI, Anthropic, any OpenAI-compatible)
- Operator Mode runs 24/7 in background detecting issues proactively
- Integrates with Kubernetes, Prometheus, Datadog, Elasticsearch, AWS, GCP, databases
- Slack integration sends fixes directly to your team
- GitHub integration can auto-open PRs to fix problems
- Terminal CLI for quick investigations from your command line
- Web interface and TUI available for non-CLI users
- K9s plugin for Kubernetes-native troubleshooting
- Automatic correlation of alerts without manual setup
- Investigates PagerDuty and OpsGenie incidents with update capability
- Combines logs, metrics, and traces in one analysis
- Provides actionable remediation steps in plain language
- Saves investigation time by eliminating manual troubleshooting
- Containerized deployment via Helm chart for Kubernetes
- REST API with streaming endpoints for real-time analysis
Holmesgpt cons
- Requires paying for external AI provider API keys (OpenAI/Anthropic costs)
- Needs proper data source configuration to access observability stack
- Older LLM models lack reasoning capability for complex problems
- Can truncate data when too much context is provided
- Vague prompts produce poor results requiring specific questions
- Requires CLI installation and AI provider setup before use
- May need company-specific runbooks customization for best results
- Local models perform worse than OpenAI for troubleshooting
- Some LLM providers reject provider-specific fields causing errors
- Kubernetes deployment requires connectivity to Prometheus/logs verified
Frequently asked questions about Holmesgpt
What is HolmesGPT?
HolmesGPT is an open-source AI troubleshooting agent built for Kubernetes and cloud-native environments. It combines observability telemetry (logs, metrics, traces), LLM reasoning, and structured runbooks to accelerate root cause analysis and suggest next actions. It was accepted as a CNCF Sandbox project in October 2025.
How do I install HolmesGPT?
You can install HolmesGPT via pip, Homebrew (brew install holmesgpt), Helm chart for Kubernetes, CLI, or UI. The easiest way is through pip or Homebrew. After installation, you must set up an AI provider API key (like export OPENAI_API_KEY) before running investigations.
What AI providers does HolmesGPT support?
HolmesGPT supports multiple AI providers including OpenAI (default), Anthropic, and any OpenAI-compatible LLM. OpenAI models provide a good balance of accuracy and speed, while Anthropic models often give better results at the expense of speed. Recommended models include anthropic/claude-opus-4-1-20250805, anthropic/claude-sonnet-4-20250514, and gpt-4.1.
What data sources can HolmesGPT connect to?
HolmesGPT automatically discovers connected data sources including Kubernetes, Prometheus, Datadog, Elasticsearch, AWS, GCP, databases, and more. It aggregates signals from observability tools such as logs, metrics, alerts, and distributed traces to perform analysis.
What is Operator Mode?
Operator Mode allows HolmesGPT to run in the background 24/7, spotting problems before customers notice and messaging you in Slack with the fix. While the operator runs in Kubernetes, health checks can query any data source Holmes is connected to including VMs, cloud services, databases, and SaaS platforms. It can also open PRs to fix what it finds when connected to GitHub.
Is HolmesGPT free?
Yes, HolmesGPT is completely free and open-source. It is a CNCF Sandbox project with no paid tiers. You only pay for the AI provider API keys you use (OpenAI, Anthropic, etc.). There are no subscription fees or enterprise plans for HolmesGPT itself.
What makes HolmesGPT different from chatbots?
Unlike static dashboards or passive chatbots, HolmesGPT is agentic: it actively decides what data to fetch, runs targeted queries, and iteratively refines its hypotheses while staying within your environment. It understands intent, creates task lists, queries data sources, correlates context, and explains with suggested fixes.
How do I get help with HolmesGPT?
You can join the HolmesGPT Slack community for help, request features or report bugs on GitHub, or open a GitHub issue for help. The community is active and members can assist with troubleshooting and setup questions.
Can HolmesGPT integrate with incident management tools?
Yes, HolmesGPT integrates with PagerDuty and OpsGenie. You can run commands like 'holmes investigate pagerduty --pagerduty-api-key <API_KEY>' to investigate incidents and use '--update' to write the analysis back to the incident as a comment. It also has Slack integration for messaging fixes directly to your team.