π PagerDuty AI Agent
@cnoe-io
PagerDuty AI Agent powered by 1st Party MCP Server using OpenAPI Codegen, LangGraph and LangChain MCP Adapters. Agent is exposed on various agent transport protocols (AGNTCY ACP, Google A2A, MCP Server)Agent for PagerDuty
Overview
What is π PagerDuty AI Agent?
An LLM-powered agent built using the LangGraph ReAct Agent workflow and MCP tools, connecting to the PagerDuty API. It supports both ACP and A2A protocols for flexible integration with external user clients, and enforces PagerDuty API token-based authentication.
How to use π PagerDuty AI Agent?
Configure a .env file with your LLM provider details and a PagerDuty API key. Pull the Docker image from ghcr.io/cnoe-io/agent-pagerduty:a2a-latest and run it on port 8000, then use the agent-chat-cli to interact. Alternatively, clone the repo and run make run-a2a for local development.
Key features of π PagerDuty AI Agent
- LangGraph + LangChain MCP Adapter for agent orchestration
- Azure OpenAI GPT-4 as the LLM backend
- Multi-protocol support (ACP and A2A)
- Comprehensive PagerDuty API coverage (incidents, services, schedules, teams, users, escalation policies)
- Secure API token-based authentication
Use cases of π PagerDuty AI Agent
- List, create, update, and resolve incidents through conversational queries
- View and manage services and their statuses
- Query schedules, teams, users, and escalation policies
- Integrate into multi-agent workflows via either ACP or A2A protocol
FAQ from π PagerDuty AI Agent
What dependencies are required?
Python 3.13+, Poetry 2.1.1+, and Docker are required. An LLM provider (Azure OpenAI in the example) and a valid PagerDuty API key are needed.
How is authentication handled?
The agent uses a PagerDuty API key set in the PAGERDUTY_API_KEY environment variable. The key is passed to the PagerDuty API for all operations.
What protocols does it support?
It supports both the ACP (AGNTCY) and A2A (Google) protocols. ACP is considered legacy; A2A is the primary mode.
Where does data live?
All data resides in the PagerDuty system. The agent acts as a client, making API calls to https://api.pagerduty.com β no local storage of PagerDuty data.
Are there any known limitations?
The README does not mention specific limits beyond requiring network access to the PagerDuty API and the chosen LLM provider.
