Robusta

What Robusta is

Robusta is an AI SRE for teams investigating infrastructure alerts and responding to incidents. It learns an environment, groups duplicate alerts, filters noise, and surfaces new issues. When an alert needs attention, its AI investigates the cause and blast radius, proposes a fix, and can escalate to a human. Alerts can arrive through webhooks, HTTP requests, or MCP events; named tools include Prometheus AlertManager, Grafana, Datadog, PagerDuty, and Jira. Holmes has native integrations for Kubernetes, OpenShift, AWS, Azure, GCP, Prometheus, Grafana, PostgreSQL, GitHub, and ServiceNow. It can connect to other data sources through MCP servers or HTTP APIs, and can generate a connector for an HTTP API. Robusta can run as SaaS, in a customer's VPC, or fully self-hosted with the customer's own LLM key. Customers can choose their own models and API keys, and disable Robusta-hosted models. Robusta says customer logs, traces, and incidents answer questions but are not used to train models. The agent is read-only by default; write access for auto-remediation is optional. Queries, log lines, and conclusions are logged and replayable. Robusta says it is SOC 2 compliant, with its report available on request. The core is open source, and pricing is provided by quote after prospective customers share team and infrastructure details.

Who it suits

Robusta suits engineering teams that monitor Kubernetes or cloud infrastructure and need help sorting alerts and investigating incidents. It may fit teams that want deployment choices, control over model keys, and an agent that starts in read-only mode.

Getting started

1. Visit https://home.robusta.dev/.
2. Choose SaaS, deployment in your VPC, or full self-hosting.
3. For a quote, provide details about your team and infrastructure.
4. Connect alert sources through webhooks, HTTP requests, or MCP events.
5. Use Robusta's offer to help install it on your cluster and walk through a live incident.

Verdict

Choose Robusta if your team needs AI assistance investigating alerts and wants control over deployment and model choice. The main consideration is that pricing requires a quote, and self-hosted deployment uses your own LLM key.