Agents · 2025
Quorum Research
A multi-agent research platform. Eight specialised agents split a research question between them, from planning and searching to fact-checking and writing, and the interface shows every step as it happens.
- Role
- Own project, from design to deployment
- Stack
- React · Python · WebSockets · AWS

01 ProblemDeep research means hours of searching, cross-checking and drafting, with scattered results.
02 ApproachEight agents each own one stage of the research, visualised live as a swarm.
03 TechnologyReact 18 and Vite, Python asyncio agents, WebSockets, Redis and PostgreSQL on AWS.
04 OutcomeA transparent pipeline from question to a branded, exportable report.
Market, academic and competitive research takes hours of manual searching, cross-referencing, fact-checking and drafting. The results end up scattered, and it is hard to see how a conclusion was reached.
01One stage per agentA Planner breaks the question down; a Researcher, an Academic Analyst and a Code Analyst gather evidence; a Fact Checker and a Critic test it; a Synthesizer and a Report Writer turn it into the final document.
02Show the reasoning, not just the answerA live orbital map shows data moving between agents, a telemetry console streams their reasoning, and a source ledger tracks trust scores, citations and relevance for every source used.
03Built to plug into any agent backendA hook-based orchestration layer keeps the interface separate from the agent framework, so it can connect to LangGraph, CrewAI or AutoGen without structural UI changes.
- Frontend
- React 18, Vite, Framer Motion
- Backend
- Python asyncio agents, Node.js APIs, WebSockets for live updates
- Data
- PostgreSQL, Redis
- Orchestration
- Hook-based layer for LangGraph, CrewAI or AutoGen backends
- Models
- Switchable reasoning models, including OpenAI and Gemini
- Infrastructure
- Docker, AWS Amplify, AWS CloudFormation
- Output
- White-label report builder with one-click PDF export
- Design
- Custom CSS design system with light and dark modes
04 — Outcome2025
agents, each owning one stage of the research
agent frameworks it can drive without UI changes
1-clickbranded PDF report from a finished run


