OpsPilot – Secure AI Agent with Auth0

An intelligent AI agent that reads emails, understands intent, and securely performs actions like replying or alerting using permission-based workflows with Auth0 and LangGraph.

Comment

Inspiration

AI agents today are powerful but unsafe. They can take actions without proper permission, which creates serious security risks. We wanted to build a system where AI can act, but safely — just like a human with controlled access.

What it does

OpsPilot is an intelligent AI operations agent that:

  • Reads incoming emails
  • Understands the context using AI
  • Classifies the type of request (security, payment, meeting, etc.)
  • Decides what action to take:
    • Auto reply
    • Send alert
    • Ask human approval
  • Executes actions like sending emails or Slack alerts

It also includes:

  • Human-in-the-loop approval system
  • Voice input and output
  • Real-time dashboard
  • Automatic report generation

How we built it

We built OpsPilot using a multi-agent architecture powered by LangGraph.

System flow:

  1. Email is fetched using IMAP
  2. Ingestion agent prepares structured input
  3. Classification agent (LLM) identifies incident type
  4. Decision agent determines next action
  5. Human agent handles approvals if required
  6. Reply agent generates smart responses
  7. Action agent executes real-world actions (email + Slack)
  8. Report agent generates logs and PDF reports

Technologies used:

  • LangGraph (multi-agent orchestration)
  • Groq + LLaMA 3 (LLM)
  • Auth0 (authentication)
  • Gradio (UI dashboard)
  • SMTP/IMAP (email)
  • Slack Webhooks
  • ElevenLabs (text-to-speech)
  • Whisper (speech-to-text)

Challenges we ran into

  • Designing a reliable multi-agent workflow
  • Ensuring proper decision routing between agents
  • Handling real-time email parsing
  • Maintaining state across multiple agents
  • Integrating voice features smoothly
  • Making the system modular and scalable

Accomplishments that we're proud of

  • Built a complete end-to-end AI agent system
  • Integrated real-world actions (email + Slack)
  • Designed human-in-the-loop control
  • Created a working multi-agent pipeline using LangGraph
  • Added multimodal capabilities (voice + UI)

What we learned

  • How to design agent-based systems using LangGraph
  • Importance of human control in AI systems
  • Real-world integration challenges with APIs
  • Building scalable and modular AI pipelines

What's next for OpsPilot – Secure AI Agent with Auth0

  • Full Auth0 Token Vault integration
  • Role-based access control
  • Cloud deployment (AWS/Docker)
  • Better UI/UX
  • Multi-user enterprise support

Built With

  • auth0
  • elevenlabs
  • gradio
  • groq-(llama-3)
  • imap
  • langgraph
  • python
  • slack-webhooks
  • smtp
  • whisper
Share this project:

Updates