notion

name	notion
description	Fetch Notion pages as Markdown from the CLI using your existing browser session cookies — no per-page integration grants. Search workspaces, list logged-in identities across browser profiles. Use when the user references a Notion page URL/ID, says "import this Notion page", asks to search Notion, or wants to inspect which Notion accounts they're signed into.

Notion CLI

Read Notion via the unofficial /api/v3 endpoint using token_v2 from the user's browser (Arc by default). Sees every page their logged-in browser sees — including pages shared by guests from other workspaces, which the official API can't reach without per-page integration grants.

Prerequisites

• User signed into Notion web in a Chromium-family browser (Arc, Chrome, Chromium, Brave, Edge, Firefox, Safari)
• uv installed — the script is PEP 723 so uv run self-resolves dependencies (browser-cookie3, diskcache, emboss, pydantic, requests, rich, typer)

Commands

All commands use notion.py from this skill dir. Default browser is Arc; override with -b chrome|brave|edge|firefox|safari|chromium.

Fetch a page as Markdown

./notion.py get <page-id-or-url>           # → stdout
./notion.py get <page-id-or-url> -o out.md # → file
./notion.py get <page-id-or-url> --raw     # raw recordMap JSON instead of rendered Markdown

<page-id-or-url> accepts:

• Full Notion URL: https://www.notion.so/Page-Title-292ff732fc3480d0a39ee6a78db70f82
• 32-hex slug tail: 292ff732fc3480d0a39ee6a78db70f82
• Dashed UUID: 292ff732-fc34-80d0-a39e-e6a78db70f82

Renders headings, paragraphs, bullets, numbered lists, code (language preserved), quotes, callouts, dividers, bookmarks, images, to-dos, toggles. Empty language=markdown code blocks (a Notion authoring quirk) are unwrapped.

List logged-in identities

./notion.py whoami

Enumerates each browser profile with a token_v2 cookie and prints the user + workspaces. Useful for figuring out which Arc profile has a particular workspace.

Search a workspace

./notion.py search "<query>" --space <space-id>
./notion.py search "<query>" --from-page <page-id>   # resolve space from a page you can access

Slack-style workspace search. --from-page is the easier path when you don't know the space ID off-hand.

Auth

Browser cookies do everything — no env vars, no tokens to manage. If a profile has Notion cookies, that profile's identity is used. With multiple profiles signed into different Notion accounts, the script uses cookies from the profile that can resolve the page.

When to use this skill

• Use when the user mentions a Notion URL/page-ID, asks to import a page into the KB, wants to scan their Notion accounts, or asks to search a workspace.
• Don't use for actions that mutate Notion (the official API + integration grants are still required for writes; this CLI is read-only).

Examples

# Import a guest-shared RunPod page into the KB
./notion.py get https://www.notion.so/Creating-an-Account-and-SSH-into-Cluster-292ff732fc3480d0a39ee6a78db70f82 \
  -o projects/anthropic-fellows/program-handbook/runpod-ssh.md

# List which Notion accounts you're signed into across Arc profiles
./notion.py whoami

# Find pages in RunPod's workspace mentioning "cluster troubleshooting"
./notion.py search "cluster troubleshooting" --from-page 292ff732fc3480d0a39ee6a78db70f82

Caching

Responses are memoized to /tmp/notion-cli-{YYYY-MM-DD} — a fresh dir per day, so cache rolls automatically. Pydantic models (NotionBlock, etc.) round-trip cleanly because emboss uses model_dump/model_validate for BaseModel returns.

Local files

• notion.py — the CLI (PEP 723 inline deps, runs via uv run)