A new RSS reader called Antenna just landed on Hacker News, and it's built specifically for AI agents. The tool stores feed subscriptions in a local SQLite database and routes new posts to both email and AI agents through MCP (Model Context Protocol). Unlike traditional readers that trap your data behind web UIs, Antenna exposes everything to software queries.

The technical setup is straightforward: Python 3.12+ on macOS or Linux, with six MCP tools that let agents list sources, search posts with FTS5 highlighting, retrieve full HTML content, and manage subscriptions. For email delivery, it supports per-post or digest modes through your own SMTP server—Amazon SES, Postmark, or Gmail app passwords all work. Deduplication uses stable entry IDs rather than URL hashes that drift.

The real innovation here is the agent-first design. Rather than bolting an API onto a human-focused reader, Antenna treats AI agents as first-class citizens alongside human users. An agent can search for posts from the last seven days tagged "robotics" mentioning "Gemini" and summarize the three most interesting—all by querying the same SQLite index your email uses.

Currently in Phase 0, Antenna ships as a source tarball to early testers under MIT license. Phase 1 will add hosted multi-user support with HTTP MCP access.