Warning
ALPHA SOFTWARE Unpage is experimental, under heavy development, and may be unstable. Use at your own risk!
Unpage is an infrastructure knowledge graph builder, and an MCP server to enable your LLM-powered application to understand and query your infrastructure.
- Python 3.12 or higher
uvpackage manager- API keys for your LLM and alerting, infrastructure, and observability tools. Learn more in Plugins.
On macOS:
brew install uvFor other platforms, follow the official uv installation guide.
Unpage is designed to be run using uvx, which comes with uv:
uvx unpage -hTo get started, run:
uvx unpage agent quickstartThis will get you up and running with your first agent, which will automatically investigate and add context to alerts from PagerDuty (or your preferred alerting provider). You will also have a chance to set up your infrastructure knowledge graph to provide your agent with more context.
The quickstart flow will walk you through:
- Configuring your LLM, PagerDuty plugin, and logs and metrics plugins
- Creating your first agent and prompt
- Testing your agent with an existing incident ticket
- Building your knowledge graph
Once you're happy with the results of your agent, you can automate the agent's actions for new incidents by running unpage agent serve and configuring PagerDuty to send webhooks to the Unpage server:
uvx unpage agent serve -hThe agent serve command supports running the server over an ngrok tunnel, so that you can test your agents end-to-end locally, without deploying. For more information on agent serve options, see its documentation.
Detailed documentation lives in docs/, and is also published via Mintlify to docs.aptible.ai.
See LICENSE.md.
Copyright (c) 2025 Aptible. All rights reserved.