Skip to main content

A real paper every morning from your private newsroom: an agent composes a print-ready edition from sources and preferences you own as files.

Project description

Morning Paper wordmark with a terminal composing a printed edition

Morning Paper

A real paper every morning, from your private newsroom.


Morning Paper turns sources and preferences you own into a finite daily edition: calm, source-backed, print-ready, and easy to mark up.

An agent composes the paper. The morning-paper CLI renders it to PDF. Your private newsroom is the folder where sources, taste, delivery rules, memory, and feedback live as files.

It is not a summarizer for a fixed feed. It is an owned algorithm you can inspect: collect candidates from many places, remember what you have already seen, filter repeats and low-signal noise, judge what deserves attention against your taste, then print an edition you can correct.

It gets better the way an editor gets better. You read the paper, give notes in natural language, and the agent turns stable feedback into durable newsroom files such as EDITORIAL.md, VISUALS.md, SOURCES.md, DELIVERY.md, and TASTELOG.md.

Set Up With An Agent

Use a strong reasoning model and paste this into Claude Code or Codex:

Install Morning Paper on this machine.

Morning Paper is an owned algorithm: an agent composes a newspaper from sources
and preferences I keep as files, and the `morning-paper` CLI renders it to a
print-ready PDF. Success is a real demo PDF on disk and open on my screen, not
just a successful package install.

Do this end to end:
1. Install the latest `morning-paper[pretty]`.
2. Confirm `morning-paper --version` matches the latest version on PyPI.
3. Run `morning-paper doctor --strict` and fix what it flags until the
   typewriter renderer and render self-test pass.
4. Run `morning-paper demo --open`.
5. If install chose the wrong Python or an old package, prefer:
   `uv tool install --python 3.13 "morning-paper[pretty]"`.

Stop when the demo PDF exists and is open on my screen. Show me the exact
commands that worked, the installed version, and the PDF path.

Do not set up my private newsroom yet. First prove the engine prints.

After the demo PDF is open, install the plugin and say: set up my morning paper.

  • Claude Code/plugin marketplace add dmthepm/morning-paper, then /plugin install morning-paper@morning-paper
  • Codexcodex plugin marketplace add dmthepm/morning-paper, then codex plugin add morning-paper@morning-paper

The plugin carries the setup, edition, and writing skills. The engine package alone renders papers; the plugin teaches the agent how to create and operate your private newsroom.

Manual Install

uv tool install --python 3.13 "morning-paper[pretty]"
morning-paper --version
morning-paper doctor --strict
morning-paper demo --open

pipx install --python 3.13 "morning-paper[pretty]" is the main fallback. Use pip only inside a virtual environment. On macOS, the print stack may also need brew install pango gdk-pixbuf.

What It Does

  • Renders a demo paper with no config or network.
  • Scaffolds a private newsroom repo with durable preferences and setup state.
  • Adds URLs or local files to a dated Assignment Board with page estimates.
  • Builds daily editions from configured sources and Assignment Board material.
  • Lets local collector scripts bring in anything else: folders, exports, Obsidian vaults, GitHub activity, business systems, research reports, or agent-produced files.
  • Typesets markdown through print style packs and chart directives.
  • Reviews a finished edition for editorial problems before delivery.
  • Works without an LLM key; the agent experience comes from Claude Code or Codex.

No database. No Docker. No hosted account required. No fabricated filler: a missing source prints "not configured" or "nothing today."

Sources

Morning Paper should meet your sources where they already live: the work that needs your attention, the people trying to reach you, the tabs you saved, the feeds you chose, the repos and tickets that changed, and the exports or folders your tools already produce. The source layer is intentionally plural: files, URLs, folders, inboxes, APIs, local scrapes, and agent-generated reports all become source markdown before the editor decides what deserves ink.

Because sources are infinite, the open-source project does not try to ship a universal adapter for every platform. It ships the newsroom contract: source ledgers, collector scripts, Assignment Board intake, memory, review, and feedback routing. A reader's private repo decides which collectors matter.

Examples:

  • email newsletters, Slack channels, Discord threads, or contributor notes;
  • GitHub activity, Linear tickets, Main Branch updates, or other work systems;
  • Twitter/X, YouTube, TikTok, Instagram, podcast, or browser-history exports;
  • folders, synced files, Obsidian vaults, saved articles, and local reports;
  • one-off URLs added to tomorrow's Assignment Board.

For social posts and threads, private collectors can write complete social source records and add them with morning-paper stage-social; snippet-only discoveries stay off the printed page until their source record is complete.

Article extraction is replaceable plumbing. The current engine has registered extractors, but the product promise is not "use this scraper." Local extraction keeps URL capture on your machine; remote readers or browser/API scrapes should be explicit choices recorded in the source item. The promise is: add source-backed material honestly, record when extraction was partial or remote, and let the editor decide what earns space.

See docs/collectors.md for the Assignment Board intake contract. See docs/source-conversion.md for small converter recipes when the local drop folder contains CSV, JSON, PDFs, vaults, work exports, or social/video history.

Recurring Editions

Prefer the native recurring primitive of the agent host you already use. Morning Paper should not invent a second scheduling system unless you ask for a local fallback.

Copy one of these into the host:

Set up a Claude Code routine that builds my Morning Paper on my chosen cadence
(weekday mornings by default unless I say otherwise). Use my private newsroom
to make the next edition and follow the Edition Run Contract: check sources,
compose, render the PDF, review, run final-editor, update memory, open or
deliver it the way my DELIVERY.md says, and tell me only if the run failed or
needs my attention. If you are in the Claude Code CLI, use /schedule to create
the routine.
Set up a Codex automation that builds my Morning Paper on my chosen cadence
(weekday mornings by default unless I say otherwise). Use this project/newsroom
to build the next edition through the Edition Run Contract, render the PDF,
update memory, and report the PDF path plus anything that needs my attention.
Set up a scheduled task for my Morning Paper. On my chosen cadence, check
whether the next edition was produced or remind me to make the paper in my
newsroom. If you have an approved way to access the newsroom runner, use it;
otherwise do not pretend you rendered the PDF.

Claude Code calls this a routine. Codex calls it an automation. ChatGPT calls it a scheduled task. The important distinction: Codex and Claude can run inside the project/newsroom; ChatGPT tasks are useful for reminders or connected workflows, but may not have access to project files and should not claim local PDF work unless the local runner is actually available. If those are not available or you specifically want a machine-local fallback, the CLI still has morning-paper routine install|status|uninstall; use it deliberately, not by default. Verify the host's current recurrence command before saving the schedule; the Edition Run Contract is stable, host scheduling commands can change.

Styles

morning-paper styles lists the print family:

Style What it is
broadsheet The newspaper you read
brief A compact paper you work through with a pen
field-card A boxed reference card
zine A half-letter photocopier handout
morning-paper render draft.md --style broadsheet --palette color

For Agents

Read AGENTS.md first. The short version:

  • agent composes, code renders;
  • use doctor --strict as proof, not vibes;
  • write setup and edition state to files before long work;
  • use edition apply-feedback to record durable reader feedback;
  • never put private newsroom facts in this public repo.

Useful CLI verbs: newsroom, sources, stage, queue, edition, estimate, render, review, local-fallback routine, doctor. sources check auto-detects a scaffolded newsroom when run from its root; use --newsroom <path> when checking from somewhere else.

Docs

Development

git clone https://github.com/dmthepm/morning-paper.git
cd morning-paper
pip install -e ".[dev,pretty]"
PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 PYTHONPATH=src python -m pytest tests/
python scripts/setup_scaffold_smoke.py --isolated
python scripts/new_user_smoke.py
python scripts/dogfood_newsroom_smoke.py
python scripts/five_edition_loop_smoke.py
python scripts/source_shape_intake_smoke.py
python scripts/host_plugin_smoke.py
python3 scripts/release_candidate_check.py --outdir /tmp/morning-paper-dist --install-check

License

MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

morning_paper-0.8.11.tar.gz (1.6 MB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

morning_paper-0.8.11-py3-none-any.whl (1.5 MB view details)

Uploaded Python 3

File details

Details for the file morning_paper-0.8.11.tar.gz.

File metadata

  • Download URL: morning_paper-0.8.11.tar.gz
  • Upload date:
  • Size: 1.6 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for morning_paper-0.8.11.tar.gz
Algorithm Hash digest
SHA256 058ddc526a32f94cbe110e70d0af1f3afdb898008b9b223fa2a3d17cbcb22f5f
MD5 ba9f744edb548c664a09a3f7b3a3ffc0
BLAKE2b-256 97fd036592a358aee5fe52855afcc1c5f4c086eeb4db1c901abea1c4705cbf41

See more details on using hashes here.

Provenance

The following attestation bundles were made for morning_paper-0.8.11.tar.gz:

Publisher: publish.yml on dmthepm/morning-paper

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file morning_paper-0.8.11-py3-none-any.whl.

File metadata

  • Download URL: morning_paper-0.8.11-py3-none-any.whl
  • Upload date:
  • Size: 1.5 MB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for morning_paper-0.8.11-py3-none-any.whl
Algorithm Hash digest
SHA256 9af1ad06549d36fc6e3013345d4d6a62a903b15db869d166a46e7ec220bd9d7d
MD5 0a3dbf05fe9723fa028166215d00a691
BLAKE2b-256 22a6321aaf25f0790a07b4fd4e284afb3c7d5e292f12fc3ce3da7eb8e1b45c61

See more details on using hashes here.

Provenance

The following attestation bundles were made for morning_paper-0.8.11-py3-none-any.whl:

Publisher: publish.yml on dmthepm/morning-paper

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page