AURA
Open-source platform for letting AI agents safely investigate and respond to production incidents, with built-in guardrails
🔗 Visit AURA
Description
Handing an AI agent access to production systems during an incident is risky — it might genuinely help, or it might make things worse with an unreviewed action at 3am. AURA is an open-source platform for deploying teams of specialist AI agents that investigate and respond to production incidents, built with production-grade safety mechanisms so agents operate within guardrails instead of having unrestricted access.
Teams configure specialist agent "crews" via TOML files, and AURA connects to observability platforms, cloud providers, and incident-management tools through the Model Context Protocol, with support for multiple LLM providers (OpenAI, Anthropic, Bedrock, Gemini, Ollama, OpenRouter) so teams aren't locked into one model vendor. It's aimed at SRE and operations teams who want AI assistance during incident response without giving an agent unrestricted access to production systems.
💬 Our review
The short version: AURA is a serious, safety-first answer to "let AI agents help during incidents" — its guardrails and state management are the actual product, not an afterthought, which matters a lot when the agents in question can touch production.
Developed by Mezmo (the observability company behind the LogDNA product already listed in this catalog), AURA is explicitly a separate, independently self-hostable open-source project rather than a bundled feature of Mezmo's core SaaS — you can run it without being a Mezmo customer. Against building your own incident-response agent from scratch, AURA's value is the pre-built safety mechanisms, multi-LLM support (OpenAI, Anthropic, Bedrock, Gemini, Ollama, OpenRouter), and MCP-based integrations into observability platforms, cloud providers and incident tools. Being free and Apache 2.0 licensed removes any cost barrier. Worth it for SRE/ops teams cautiously experimenting with AI-assisted incident response; skip it if you're not ready to let an agent near production systems at all, guardrails or not.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Open source sous licence Apache 2.0, gratuit, auto-hébergé.
Pros
Garde-fous de sécurité production intégrés
Support multi-LLM (OpenAI, Anthropic, Bedrock, Gemini, Ollama, OpenRouter)
Intégrations via Model Context Protocol vers outils d'observabilité/cloud/incident
Gratuit, open source (Apache 2.0), indépendant du SaaS Mezmo
Cons
Configuration via TOML nécessite un vrai travail de setup
Sujet sensible par nature (agents IA proches de la prod)
Développé par Mezmo — dépendance de fait à leur roadmap open-source