Skip to main content

Getting Started

Under 5 minutes. Sign up, add a connector (URL + API token), and ask your first question. See the Quick Start guide.
No. Foggy reads from your existing tools — it doesn’t replace them. Think of it as an AI SRE that works on top of your current observability stack. You keep using Grafana, Prometheus, and Loki as you always have. Foggy just queries them to answer your questions.
Currently: Grafana (includes Prometheus and Loki data sources), Kubernetes, Slack, Grafana Alerts, and Alert Manager. See our Integrations page for the full roadmap.

AI & Models

Claude Sonnet 4.5 (default), Claude Opus 4.6, and other models. You choose per-message via the model selector in the chat input.
Investigation prompts and tool query results (metrics, logs, events) are sent to the selected LLM for analysis. Infrastructure credentials (API tokens, URLs) are never sent to LLMs. See Security for the full data flow.
Simple queries: 10–30 seconds. Complex multi-tool investigations: 1–3 minutes. Foggy streams results in real-time so you see progress immediately — you don’t wait for the full answer.
Yes. BYOK (Bring Your Own Key) support lets you use your own LLM API keys. You control which LLM provider processes your data.

Security & Data

Foggy operates in read-only mode with least-privilege access and processes data only during active investigations. We are building with SOC 2 and GDPR requirements in mind. See the Security page for full details.
Never. Foggy uses read-only access to all connected tools. It queries dashboards, metrics, and logs but never creates, updates, or deletes resources in your infrastructure.
Encrypted at rest using Fernet symmetric encryption. Decrypted only when making API calls to your tools. Never logged, cached, or displayed in the UI after initial entry.
No. Foggy does not store raw production data or use it to train models. Data is processed for each investigation and retained only as conversation history within your project. Your data is never shared across projects or used for any other purpose.