I built OmniAgent because I kept running into the same wall: AI coding assistants are great at answering questions, but they can't actually do anything. They suggest. You implement. OmniAgent flips that .
It's a platform where multiple specialized AI agents work together — one writes code, one reviews it, one debugs it, one thinks about the architecture — and they hand work off to each other in a real pipeline. The idea is that you describe a task, and the agents figure out how to tackle it from every angle without you babysitting the process.
🛠️ What it does:
Coding Agent — Takes a description of what you need and writes working code. You give it a problem statement, it produces files.
Review Agent — Analyzes code diffs and pull requests. Grades issues by severity and gives you actionable suggestions.
Debug Agent — You paste in code and an error, it traces the failure, identifies the root cause, and produces a fix.
Architect Agent — Generates project scaffolds, writes architecture decision records, and produces Mermaid diagrams.
Multi-Agent Pipelines — Chain agents together: architect designs the approach, coder implements it, reviewer checks it, debugger fixes what breaks.
Real-Time Streaming — Everything streams over WebSocket. You watch the agent think and write in real time.
Repo Analysis — Point it at a local git repository and it'll parse the structure, understand the language breakdown, and answer questions about the codebase.
🧰 Tech Stack
Backend is Python with FastAPI streaming WebSocket responses. AI providers are OpenAI and Anthropic — you can switch between them. Frontend is React 18 with TypeScript and Tailwind CSS. PostgreSQL stores task history and Redis handles caching.
🔗 Get Started
Check out the repo: github.com/isidhartha/omni-agent
Top comments (0)