Quick overview
Reads Network report queries and agent prompts from GitHub, uses AI Agents with a read-only Cisco pyATS MCP tool to generate infrastructure audit reports, automatically opens GitHub issues for detected problems, and sends Slack notifications with report and issue links. Demo: http://cs.co/9006BEWHQy Repo: http://cs.co/9007BEmQdZ
How it works
- Runs on a schedule or when executed manually.
- Pulls a list of report query
.txt files from a GitHub repository and iterates through them one by one.
- For each query, fetches the query file and the reporting agent prompt from GitHub
- Uses a Reporting AI Agent with a read-only Cisco pyATS MCP tool to query network devices and generate a structured report.
- Saves the generated report as a Markdown file in GitHub and extracts the report URL and tag.
- Uses a ticketing AI Agent to decide whether the report requires a ticket, then creates a GitHub issue when the decision indicates an issue.
- Sends a Slack message card with the report link and, when applicable, the GitHub issue link, then continues to the next query file.
Setup
- Add GitHub credentials and set the repository owner/name plus the file paths for the report queries folder, the reporting prompt file, the ticketing prompt file, and the reports output folder.
- Add your LLM API credential of choice for both AI Agents.
- Configure the read-only Cisco pyATS MCP endpoint URL and ensure the required pyATS tools are enabled and can reach your network devices.
- Add Slack credentials and replace the target channel ID in both Slack message steps.
- Populate your GitHub repo with report query
.txt files formatted as key/value lines (including a file_prefix) and adjust the schedule interval to your desired cadence. Check the green post-it in the workflow for examples.
Requirements