Operate your network directly from Slack using a multi-agent architecture.
Highlights
- 💬 Slack-based ChatOps interface
- 🧠 Multiple specialized agents (intent, planning, read, commit, formatting)
- 🛡️ Guardrails with human approval before writes
- 🔌 Real execution via pyATS MCP
- 🔁 Rollback-aware configuration workflows
Use cases
- Safe interactive automation
- Change workflows with approval
- Secure self-hosted NetOps bots
📁 More information:
Agentic ChatOps for network automation using n8n + Slack + pyATS MCP
An autonomous pipeline for continuous network reporting and GitHub-native remediation tracking.
Highlights
- 📅 Scheduled or manual execution
- 🤖 Reporting agent investigates network using pyATS MCP
- 📄 Generates professional Markdown reports
- 📁 Commits reports to GitHub
- 🎫 Ticketing agent detects risks and auto-creates GitHub Issues
- 🔗 Tight traceability between evidence and action
Use cases
- Continuous posture reporting
- Compliance evidence
- Risk-driven backlog generation
- GitOps-style operational governance
📁 More information:
Agentic reporting + GitHub issue automation using n8n + pyATS MCP
Clone the pyATS MCP repository:
git clone https://github.com/ponchotitlan/pyATS_MCP
Follow the instructions in the README of this repository to setup the MCP server in HTTP mode. Afterwards, clone this repository:
git clone https://github.com/ponchotitlan/pyATS-loves-agenticops
Open the docker-compose.yml file and update the following parameters of the services:
# Webhook URL - Get your static domain from https://dashboard.ngrok.com/cloud-edge/domains
# Then replace YOUR-STATIC-DOMAIN below with it (e.g., abc-123-def.ngrok-free.app)
- WEBHOOK_URL=uYOUR-STATIC-DOMAIN
ngrok:
. . .
- "--domain=YOUR-STATIC-DOMAIN"
# Get your auth token from https://dashboard.ngrok.com/get-started/your-authtoken
- NGROK_AUTHTOKEN=YOUR-TOKEN
Execute the following command to create a ngrok and n8n pair of containers:
docker compose up -d
To stop the services, issue the following command:
docker compose down
| Principle | Meaning |
|---|---|
| 🧩 Separation of concerns | No agent has unlimited power |
| 🔒 Guardrails by design | Writes require structure + approval |
| 🔧 Tools over hallucinations | Real data via pyATS MCP |
| 📁 Git as source of truth | Prompts, reports, tickets are versioned |
| 👤 Humans stay in control | Automation proposes, humans approve |
Assume models are fallible. Design systems that remain safe anyway.
Both projects rely on a shared real-world execution backend:
This provides:
- Real CLI execution
- Real device outputs
- Zero fabricated telemetry
- Clean separation between reasoning and execution
- pyATS – network interaction engine
- MCP server – execution abstraction
- n8n – workflow orchestration
- LLM agents – reasoning layer
- GitHub – storage, audit trail, collaboration
- Slack – ChatOps interface (project 1)
- Agentic NetOps experimentation
- Secure automation architecture
- Internal platform foundations
- Operational AI
- GitOps + AI convergence
Got ideas for new workflows or improvements? Contributions are welcome! Feel free to open issues or submit pull requests.
Made with ☕️ by Poncho Sandoval -
Developer Advocate 🥑 @ DevNet - Cisco Systems 🇵🇹
