Hyground vs
Robusta / HolmesGPT
Hyground vs Robusta / HolmesGPT: governed AI operations inside your own cluster
HolmesGPT is an open-source investigation agent: read-only, LLM-agnostic, with an optional SaaS platform from Robusta for history and change tracking. Hyground runs on the same architecture (in-cluster, bring your own LLM, read-first by default) and ships it as a governed, supported platform, which means one central gateway with platform-level credential management and a full audit trail, plus workflow automation across your operations.
A fair starting point
HolmesGPT is a CNCF Sandbox project, created by Robusta, with Microsoft contributing. It runs an agentic loop over your observability toolsets, stays read-only, and respects RBAC. You can run it as a CLI, as an HTTP API, or as an in-cluster operator, and the Operator mode works in the background: it schedules recurring health checks and opens pull requests to fix what it finds. The richer history, change tracking, and dashboards sit in Robusta's SaaS platform. Hyground takes the same stance, in-cluster, bring your own LLM, read-first. What it adds is governance: one gateway, credentials managed at platform level rather than on laptops, an audit trail of every AI interaction, and supported connectors across your observability tools and your ITSM tools.
Side by side
Hyground vs
Robusta / HolmesGPT
at a glance
What matters
Hyground
Robusta / HolmesGPT
Where it runs
In your own infrastructure: any conformant Kubernetes cluster, in your own cloud or on-premises, up to fully air-gapped.
In your own cluster, as a CLI, as an HTTP API, or as an operator. History and dashboards live in Robusta's SaaS platform.
What you operate
A supported product with vendor support and a published version-support policy.
An open-source project that you host, integrate, and operate yourself.
Governance and audit
One in-cluster gateway, credentials managed at platform level, and a full audit trail of every AI interaction.
Kubernetes RBAC per service account. You build cross-tool governance and a unified audit trail yourself.
LLM choice
Any LLM provider, through LiteLLM: cloud models in your own tenant, self-hosted models, or any OpenAI-compatible API. No fixed list.
Anthropic, Amazon Bedrock, Azure AI Foundry, Gemini, Google Vertex AI, OpenAI, Ollama, or a LiteLLM proxy.
Recurring and background work
Scheduled investigations, cron or one-time, and reusable skills. Every run goes through the same gateway and the same audit trail.
Operator mode runs in the background, schedules health checks through a Kubernetes resource, and drafts pull requests with fixes.
ITSM and connectors
Supported Jira and ServiceNow connectors with gated write-back: Jira issues and comments, ServiceNow incident comments.
Alerts from PagerDuty, OpsGenie, and Jira with findings written back, through the CLI. ServiceNow read toolset. Jira via Atlassian Rovo MCP.
Certification
ISO/IEC 27001:2022 certified.
A CNCF Sandbox project. Robusta states SOC 2 compliance for its own platform.
License and cost
Priced on infrastructure size, not seats. Quote on request.
Free and open source. Robusta's platform is priced on request.
Why teams choose Hyground
Where Hyground differs
Decision
When each platform fits
HolmesGPT and Hyground share an architecture. They sit at different points on the build-versus-buy line: HolmesGPT is an open-source agent, and Hyground is a governed platform.
Choose
Robusta / HolmesGPT
when
You want a free, open-source troubleshooting agent, you have the team to run it and integrate it, and you are willing to add Robusta's SaaS platform for the history and change tracking.
FAQ
Hyground vs
Robusta / HolmesGPT
:
common
questions
Is Hyground an alternative to HolmesGPT?
Yes, when you want the same approach as a supported platform. Both investigate inside your own cluster with a model that you choose. HolmesGPT is an open-source agent that you operate yourself, and Hyground adds a central gateway, platform-level credential management, an audit trail, ITSM connectors and vendor support.
What does Hyground add to the HolmesGPT approach?
Governance and support. Every tool call goes through one gateway in your cluster, credentials are managed at platform level instead of on laptops, and every AI interaction is logged. Hyground also schedules full investigations, ships supported Jira and ServiceNow connectors, and comes with a vendor and ISO/IEC 27001:2022 certification. With HolmesGPT, governance across tools is yours to build.
Can Hyground run air-gapped or EU-only?
Yes. Hyground deploys into your own infrastructure, from your own cloud to a fully air-gapped site, and there is no data egress to Hyground as the vendor. With a self-hosted model, no data leaves your infrastructure at all.
Can we choose the model, or host it ourselves?
Yes. Hyground connects to any LLM provider through LiteLLM, including a cloud model in your own tenant or a model that you host yourself. HolmesGPT is LLM-agnostic too. With Hyground, the model sits behind one gateway, with platform-level credentials and a full audit trail.
How do we get Hyground running, and what does it cost?
A forward deployed engineer works with your team from install to daily use. The engineer connects your stack, builds any connector that you are missing, and runs the training and workshops. Hyground is priced on the size of the infrastructure that it covers, not per seat, and you get a quote on request. HolmesGPT is free, and its cost is the engineering time that it takes to run.