← Back to work
Project Case Study

Brain+

Multi-agent Jira copilot for product operations, research, and grounded execution

A multi-agent Jira copilot built for real product operations work: backlog refinement, issue creation, document drafting, project memory, RAG over internal docs, live research, and browser-side capture through a Chrome extension.

FastAPILangGraphReact 18TypeScriptChrome Extension MV3GeminiChromaDBPydantic
Status
Active AI product
Timeline
2026 · active full-stack AI product
Links
Metrics
Web app, Chrome extension, and streaming API
Interaction surfaces
Router, Jira, RAG, Research, Notion, and Responder
Agent roles
Project memory + local vector retrieval
Grounding model
01
Overview

Brain+ combines Jira execution, grounded retrieval, live research, and project memory for backlog work, discovery tasks, and document drafting.

02
Key Highlights
  • LangGraph routes work between specialized agents instead of relying on one giant prompt, separating Jira actions, RAG lookup, research, and response synthesis.
  • The product ships across multiple surfaces: a React web app for deep work, a Chrome extension overlay for in-context capture, and a streaming backend for longer-running flows.
  • Project memory and a local vector store ground answers in prior context and uploaded documents.
  • Browser-side capture makes the assistant usable inside real workflows by pulling page context, images, and voice input directly from the current tab.
03
Technical Notes
  • FastAPI exposes the streaming API layer while LangGraph coordinates routing, tool execution, and agent-to-agent handoffs.
  • React 18 + TypeScript power the main product UI, including project-aware chat, markdown rendering, document handling, and settings flows.
  • The Chrome extension uses Manifest V3, Shadow DOM isolation, storage sync/local, and rich capture inputs to stay useful on any page.
  • ChromaDB and project memory provide the retrieval layer, while Gemini handles reasoning and search-grounded research workflows.
04
Results
  • Combines web app, Chrome extension, and streaming API around one agent graph.
  • Separates Jira actions, retrieval, research, and response generation into clear agent roles.
  • Shows full-stack AI product work across orchestration, UX, retrieval, and integrations.
← PreviousWakeTheBookNext →Dopóki Życie Trwa