Skip to main content

Your infrastructure.
Your rules.
Your SRE agent.

Accelerate SRE work with AURA, an open source agent that runs in your environment with your tools, telemetry, and operational context. Install it in one command and point it at a live incident.

Get started with AURAView source on GitHub
# 1. Install the CLI
# 2. Generate config.toml (provider, key, model)
$ aura init
# 3. Start the agent and connect a tool over MCP
$ aura
❯ /mcp add
# 4. Put it to work
❯ investigate the recent payment failures

Choose your path

Two products, documented separately. Switch between them at any time from the product menu at the top of the page.

AURA

Open source, Apache 2.0, written in Rust. An SRE agent harness you declare in TOML and run on your laptop, in Docker, or in Kubernetes. Scoped tool access per worker, approval gates on writes, and OpenTelemetry spans for every model and tool call.

Mezmo Platform

Commercial telemetry platform. Log Management, Telemetry Pipelines, Mezmo Edge, and the Mezmo API for collecting, optimizing, and routing logs, metrics, and traces at production volume.

Get started with AURA

From install to your first investigation in four steps. You need macOS or Linux and an API key for OpenAI, Anthropic, Gemini, Bedrock, or OpenRouter, or a local Ollama instance.

1

Install the CLI

The script picks a native package, Homebrew, or a direct binary for your platform. See Installation for every method.
2

Generate a config

The wizard detects your provider key, offers a model shortlist, and writes a config.toml for an SRE orchestrator with three specialist workers. Your key stays in the environment, never in the file.
3

Connect your tools

Start AURA, then add an MCP server from the built-in catalog (Mezmo, PagerDuty, Datadog, Kubernetes) or any server over HTTP, SSE, or stdio. You choose which workers see which tools.
4

Put it to work

Restart aura to load the new server, then ask it something real: Who is on call right now, and are there any open incidents? The full walkthrough is in the Quickstart.

Local quickstart

Install the binary, run aura init, connect an MCP server, and start asking questions.

Docker quickstart

The full stack with docker compose: orchestrator mode, a chat UI, and a trace viewer.

Kubernetes quickstart

A long-lived in-cluster SRE agent with Kubernetes and Prometheus MCP servers.

How AURA stays yours

You own the config, the deployment, and the telemetry. Nothing has to route through a vendor for the agent to work.

Your infrastructure

Run on a laptop, in Docker, or in-cluster. Swap between OpenAI, Anthropic, Gemini, Bedrock, OpenRouter, or local Ollama with one provider block.

Your rules

Each worker gets an explicit tool list. Approval gates name the mutating tools that wait for a human. Reads run straight through.

Your SRE agent

Declare workers, prompts, runbooks, and skills in TOML. Every model and tool call emits an OpenTelemetry span you can read in any backend.

Mezmo Platform

Already running Mezmo for telemetry? Jump straight to the product you use.

Log Management

Collect logs, parse fields, search events, create views, and alert on change.

Telemetry Pipelines

Receive, process, route, and deliver telemetry to the destinations you choose.

Mezmo Edge

Run pipelines inside your own network, close to your sources.

API Reference

Authenticate once, then script Log Analysis and Pipeline workflows.

Build AURA in the open

Ask questions, share the config you are running, and talk through an approach before you open a PR.

Join us on SlackGet started on GitHub