Inspiration
Classic RSS and Atom feeds formed the open backbone of the decentralized web, but in today's fast-moving software ecosystem, feeds often overwhelm developers with sheer volume. We built GhostFeed to resurrect this timeless syndication technology with modern AI agent workflows.
What it does
GhostFeed is an intelligent web feed companion that monitors RSS/Atom subscriptions, automatically cluster-tags articles, synthesizes technical highlights, and lets users converse with feed content in natural language using Kiro-driven agents and custom tools.
How we built it
- Core Engine: Developed using Kiro IDE and powered by Claude and GPT models.
- Agent Integration: Leveraged Kiro custom hooks, steering docs, and MCP servers to connect real-time RSS stream parsers with LLM evaluation pipelines.
- Frontend: Clean and responsive web dashboard crafted for fast triage and semantic reading.
Challenges we ran into
- Handling disparate feed schemas and malformed XML timestamps across legacy endpoints.
- Designing low-latency semantic caching so conversational queries return immediate answers without re-fetching full article histories.
Accomplishments that we're proud of
- Seamless re-animation of legacy RSS sources into structured, interactive knowledge streams.
- Zero-friction developer experience accelerated by Kiro's AI coding capabilities.
What we learned
- How to orchestrate agent steering and custom MCP servers effectively within Kiro.
- The enduring power of decentralized web standards when paired with cutting-edge LLMs.
What's next for GhostFeed
- Integration with desktop notifications and webhooks for breaking tech alerts.
- Community-curated feed directories and multi-agent debate synthesis.
Built With
- claude
- gpt
- kiro
- mcp
- node.js
- react
- rss
- typescript
Log in or sign up for Devpost to join the conversation.