Quickstart¶
Get Distill running in 5 minutes. All you need is Python and an Anthropic API key.
Prerequisites¶
| Requirement | Version | Install |
|---|---|---|
| Python | 3.11+ | python.org |
| uv | latest | curl -LsSf https://astral.sh/uv/install.sh \| sh |
| Bun | latest | curl -fsSL https://bun.sh/install \| bash (for web dashboard) |
Step 1: Clone and install¶
This installs all Python dependencies including trafilatura (article extraction), feedparser (RSS), and the anthropic SDK.
Step 2: Set your API key¶
Distill uses the Anthropic Python SDK directly. The default model is claude-sonnet-4-6.
Step 3: Run your first digest¶
This will:
- Fetch articles from 90+ built-in RSS feeds (tech, AI, engineering)
- Extract full text from each article using trafilatura
- Auto-tag and classify content
- Send everything to Claude for synthesis
- Write a markdown digest to
./insights/intake/YYYY-MM-DD.md
First run takes 2-3 minutes
Most time is spent fetching RSS feeds and extracting article text. The LLM synthesis itself takes ~30 seconds.
Step 4: Read your digest¶
Open the generated file:
Or start the web dashboard:
Then open http://localhost:6107 in your browser.
Step 5: Add your own feeds¶
Create .distill.toml in the project root:
[user]
name = "Your Name"
role = "software engineer"
[intake]
use_defaults = true # keep the built-in 90+ feeds
rss_feeds = [
"https://your-favorite-blog.com/feed",
"https://another-blog.com/rss.xml",
]
Run again:
What's next¶
- Automate it: Set up a daily cron so digests appear every morning
- Web dashboard: Start the dashboard to browse digests
- Share from phone: Set up iOS Shortcut to share articles
- Customize: Full configuration reference
Full pipeline (optional)¶
If you also use Claude Code or Codex CLI, Distill can generate a dev journal from your coding sessions and synthesize blog posts:
# Everything: sessions + journal + intake + blog
uv run python -m distill run --output ./insights --use-defaults --global
The --global flag scans ~/.claude and ~/.codex for session data.