Fusion: Your Unified Project Management Co-pilot for Software Engineering Teams
Inspiration
Project managers overseeing software engineering teams face the daily challenge of juggling multiple tools: Jira for task tracking, GitHub for code reviews, Notion for documentation, and Slack for team communication. Each platform represents a siloed ecosystem, forcing managers to constantly switch contexts and disrupting workflow efficiency. Fusion was created to transform how project managers coordinate and streamline their teams by providing a single, intelligent conversational interface that unifies all these critical tools.
What It Does
Fusion is an AI-powered co-pilot designed specifically for project managers to unify their entire software development management stack into a single chat interface. It empowers managers to use natural language to create, track, and update tasks across platforms effortlessly. Key capabilities include managing sprints, monitoring pull requests, updating documentation, and keeping team communication centralized, all without disruptive context switching. Fusion provides real-time insights such as identifying blocked issues, showing the latest commits, or summarizing ongoing workstreams. It also enables direct messaging, file sharing, and automates multi-step workflows like linking Jira issues with GitHub pull requests and syncing relevant Notion pages for comprehensive project coverage.
How We Built It
The backend infrastructure leverages Node.js and Express to maintain a resilient command center for workflow orchestration. At its core, Google’s Gemini 2.5 Pro powers the AI engine, enabling advanced natural language understanding and dynamic function chaining across services. The system architecture follows a modular MCP paradigm, where each integration (Jira, Notion, GitHub, and Slack) is represented as an independent “Tool.” We integrated CodeRabbit through GitHub to enable automated code reviews via our MCP service, and incorporated Poke to send summaries and reminders via SMS. On the frontend, we use React, TypeScript, and Vite, complemented by Tailwind CSS, to deliver a clean, responsive UI that allows project managers to interact intuitively and efficiently.
Challenges
Harmonizing the different behaviors of multiple platforms required developing a robust set of tools for our AI to use. Creating a clean, modular MCP-inspired architecture was essential for integrating all the functionalities cohesively, enabling seamless communication between each service tool while maintaining scalability. Additionally, ensuring secure handling of things like permission controls and synchronization integrity across various services required careful system design and constant testing.
Accomplishments
Fusion autonomously executes complex, multi-service workflows, such as creating and linking Jira issues from Notion documentation while simultaneously opening corresponding GitHub pull requests, significantly reducing manual coordination. The MCP-inspired modular backend architecture allowed quick onboarding of different services with minimal disruption. Example runs showed Fusion effectively managing real-world software projects, even helping us orchestrate our own project and divide tasks and issues cleanly through natural language.
Lessons Learned
Clear and precise definitions of what each tool can do make a big difference in how reliably the AI can coordinate actions across multiple services. Investing early in a modular architecture that keeps dependencies separate really pays off when it comes to scaling and maintaining the system as the project grows. Lastly, having detailed and logically placed print statements as well as catching errors can significantly speed up development and reduce future issues.
Future Directions
Planned enhancements include interactive approval workflows enabling managers to review AI-proposed changes before execution, integrations with Figma, Sentry, and CI/CD ecosystems, and RAG to enhance chatbot memory and semantic searching over chat history.
Fusion is made to empower project managers by simplifying the complexity of software team management. By unifying disjointed tools into a natural language-driven interface, Fusion restores focus, boosts productivity, and redefines how software projects are managed end to end.
Built With
- coderabbit
- express.js
- gemini
- github
- jira-api
- mcp
- node.js
- notion-api
- poke-mcp
- react
- slack-api
- tailwindcss
- typescript/javascript


Log in or sign up for Devpost to join the conversation.