CLI Tool

Everything the MCP server can do, from your terminal. Read feeds, search stories, manage subscriptions, and train classifiers. Pipe to jq, script your workflows, or just browse your feeds.

Looking for the AI agent integration? See the MCP Server →

Get started

Install

pip install newsblur

Login

newsblur auth login

Opens your browser to authorize via OAuth. Token stored locally at ~/.config/newsblur/.

Read stories

Get unread stories from feeds, folders, or everything at once.

newsblur stories list                          # unread stories
newsblur stories list --folder Tech --limit 5  # filter by folder
newsblur stories search "machine learning"     # full-text search
newsblur stories saved --tag research          # saved stories by tag
newsblur stories infrequent                    # rarely-publishing feeds
newsblur stories original 123:abc456           # fetch full article text

Daily briefing

Get an AI-curated briefing with top stories, infrequent site gems, and long reads.

newsblur briefing                              # today's briefing
newsblur briefing --limit 1                    # just the latest
newsblur briefing --json                       # structured output

Manage feeds and folders

Subscribe, unsubscribe, and organize your feed list.

newsblur feeds list                            # all subscriptions
newsblur feeds folders                         # folder tree with counts
newsblur feeds add https://example.com         # subscribe
newsblur feeds add https://blog.com -f Tech    # subscribe into a folder
newsblur feeds remove 42                       # unsubscribe
newsblur feeds organize move_feed --feed-id 42 --from News --to Tech

Take actions on stories

Save, share, and mark stories as read.

newsblur save 123:abc --tag ai --tag research  # save with tags
newsblur unsave 123:abc                        # remove from saved
newsblur read --feed 42                        # mark feed as read
newsblur share 123:abc --comment "Worth reading"

Train intelligence

View and update your classifiers to like or dislike authors, tags, titles, and feeds.

newsblur train get --feed 42                   # view current training
newsblur train like --feed 42 --author "Name"  # train a like
newsblur train dislike --feed 42 --tag sponsor # train a dislike

Discover new feeds

Search by topic, find similar feeds, or browse what's trending.

newsblur discover search "machine learning"    # search by topic
newsblur discover similar --feed 42            # find similar feeds
newsblur discover trending                     # trending feeds

Scripting and automation

Every command supports --json for structured output and --raw for unformatted text. Pipe to jq, use in shell scripts, or connect to self-hosted instances.

# Pipe to jq for structured processing
newsblur stories list --json | jq '.items[].title'
newsblur briefing --json | jq '.items[0].section_summaries'

# Use with a self-hosted NewsBlur instance
newsblur --server https://my-newsblur.example.com auth login

# Script common workflows
newsblur stories list --folder Tech --json | \
  jq -r '.items[] | "\(.title) - \(.feed_title)"'

How does authentication work?

Running newsblur auth login opens your browser for OAuth authorization, the same flow used by third-party apps. Your token is stored locally at ~/.config/newsblur/ and never sent anywhere except the NewsBlur API.

What plan do I need?

CLI access requires a Premium Archive ($99/year) or Premium Pro ($29/month) subscription.

Can I use it with a self-hosted instance?

Yes. Pass --server https://your-instance.com to any command, or set the NEWSBLUR_SERVER environment variable.

What's the difference between CLI and MCP?

The CLI is a standalone terminal tool you run directly. The MCP server lets AI agents like Claude and Codex connect to your NewsBlur on your behalf. Both use the same underlying API and have the same capabilities.