RSS feed reader with AI relevance scoring. Fetches feeds, scores articles against your interest profile using Claude, and surfaces what matters.
Download a prebuilt binary from Releases:
# macOS Apple Silicon
curl -L https://github.com/mcyork/feed/releases/latest/download/feed-darwin-arm64 -o feed
chmod +x feed && mv feed /usr/local/bin/
# macOS Intel
curl -L https://github.com/mcyork/feed/releases/latest/download/feed-darwin-x64 -o feed
# Linux x64
curl -L https://github.com/mcyork/feed/releases/latest/download/feed-linux-x64 -o feed
# Linux ARM64
curl -L https://github.com/mcyork/feed/releases/latest/download/feed-linux-arm64 -o feedVerify checksums against SHA256SUMS in the release.
git clone https://github.com/mcyork/feed.git
cd feed
bun installANTHROPIC_API_KEYenvironment variable (for scoring) — add to.envin the project root
# Binary
feed manage add https://workshop.esoup.net/feed.xml
feed manage add https://www.schneier.com/feed/atom security
feed manage add https://tldrsec.com/feed.xml security
# From source
bun run src/index.ts manage add https://workshop.esoup.net/feed.xmlfeed fetch
# or from source: bun run src/index.ts fetchFetches all subscribed feeds, deduplicates articles, and scores new ones against your interest profile using Claude Haiku. Supports conditional requests (ETag/304) to skip unchanged feeds.
feed surface # Score 7+ (default)
feed surface 5 # Score 5+Articles scoring 9-10 are flagged as "must read."
feed manage list
feed manage remove https://example.com/feed.xmlYour profile lives at ~/.config/feed/profile.json. Edit it to tune scoring:
{
"description": "Software engineer interested in the edge of technology",
"topics": [
"security",
"DNS and internet infrastructure",
"cryptocurrency and programmable money",
"AI agents and coding assistants",
"TypeScript and Bun"
]
}All state is stored in ~/.config/feed/:
| File | Purpose |
|---|---|
feeds.json |
Subscriptions |
items.jsonl |
All fetched articles with scores |
state.json |
Fetch timestamps, ETags |
profile.json |
Interest profile for scoring |
This project includes a Claude Code skill for conversational feed management. Install by copying the skill/ directory to ~/.claude/skills/Feed/.
The skill is cryptographically signed with SkillSeal. Verify with:
skillseal verify ~/.claude/skills/FeedSee LLMS.md for machine-readable project documentation.
MIT