What it does

ClawFeed ingests sources—Twitter feeds, RSS, Hacker News, Reddit, GitHub Trending, even arbitrary websites—and generates structured summaries on a fixed cadence: every 4 hours, daily, weekly, or monthly. The output is a clean web dashboard plus shareable RSS/JSON feeds. It runs standalone via Docker or npm, or drops into OpenClaw and Zylos as a skill.

The interesting bit

The “Source Packs” feature lets users bundle and share curated source collections, which turns the tool into something closer to a community-curated newsletter platform than a personal reader. Also notable: it uses SQLite as the default store, which keeps deployment friction low, and the README documents a full Caddy reverse-proxy setup—rarely seen in hobby projects.

Key highlights

• Multi-frequency digests (4h / daily / weekly / monthly) with configurable AI prompts and curation rules via markdown templates
• 9 source types including Twitter lists, subreddits, GitHub trending, and custom JSON APIs
• Per-user RSS and JSON Feed output, so others can subscribe to your digest
• Google OAuth multi-user support; runs read-only without it
• 66 E2E tests included in the repo

Caveats

• The README mentions “AI-powered” throughout but never names the model, API, or cost structure—whether it calls OpenAI, runs local LLMs, or something else is unclear
• “Smart curation” and “feed quality analysis” are described as features, but the actual mechanics (ML classifier? heuristic rules? prompt-based filtering?) are not specified
• Twitter/X integration may face stability issues given the platform’s API volatility, though this is not acknowledged in the docs

Verdict

Worth a look if you want a self-hosted alternative to Briefing or Mailbrew with more source flexibility. Skip it if you need transparency about the AI backend or are looking for a managed SaaS with predictable pricing.