Thread Lens

Multi-agent deep research, sourced from Reddit.

---

How it works

Step	What happens
Clarify	Follow-up questions sharpen your query
Orchestrate	A planner agent breaks it into sub-tasks
Research	Parallel sub-agents search Reddit posts and subreddits
Synthesize	A final agent consolidates findings into a structured report
History	All research is saved and browsable

Stack

Layer	Tech
Frontend	Next.js, React, TypeScript, Tailwind CSS
Backend	FastAPI, LangGraph
Reddit access	MCP server (packages/mcp-reddit)
Database	SQLite (packages/db)

Getting started

Docker (recommended)

cp .env.example .env  # fill in your API keys
docker compose up

Frontend → http://localhost:3000 · Backend → http://localhost:8000

Manual

Backend

cd apps/research
uv sync
uv run uvicorn research.main:app --reload

Frontend

cd apps/client
npm install
npm run dev

Environment variables

See .env.example. You need at least one GOOGLE_API_KEY_* for the research agents.