PratsPilot is a full-stack Agentic AI mission-control platform for creating, orchestrating, and evaluating AI agents without writing agent code.

Users can define an agent through its goal, role, instructions, expected output, model, knowledge documents, and tools. Agents can then be connected into multi-step workflows where structured outputs move between steps.

Core capabilities:

- Create specialized AI agents without coding them

- Use Gemini, Groq, Cerebras, and OpenRouter models

- Attach PDF, DOCX, Excel, Markdown, JSON, and text knowledge

- Create custom Python tools with securely stored secrets

- Equip agents with live web search and external API tools

- Build visual multi-agent workflows

- Pass mapped JSON state between workflow steps

- Insert Human Approval Gates at selected stages

- Generate agents and workflows through Autopilot

- Monitor outputs, logs, latency, tokens, and execution history

- Challenge other creators in an AI-judged Agent Battle Arena

- Separate shared User and restricted Admin realms

Built with .NET 10, React, Material UI, PostgreSQL, Python, Docker, Vercel, and Render.

Live platform: https://pratspilot.vercel.app

Source code: https://github.com/Prats222/Agentic_AI_Platform

Testing note: the free backend may require around 30 seconds to wake after inactivity.