saas-scraper 0.3.0

Chrome-driven SaaS content scraper — yields a uniform Document stream for downstream pipelines (pleno-anonymize, pleno-secret-scanner).

saas-scraper

Chrome-driven SaaS content scraper. Yields a uniform Document stream for downstream pipelines (e.g. pleno-anonymize, pleno-secret-scanner).

Where API-based connectors stop — locked-down workspaces, SSO-only sessions, content only visible in the UI — saas-scraper keeps going by driving a real Chrome session via Playwright. Reuses your existing browser profile so login, MFA and SSO flows are inherited rather than re-implemented per provider.

Install

uv add saas-scraper
# one-time browser binary install
uv run playwright install chromium

Or as a CLI:

pipx install saas-scraper
playwright install chromium

Usage

# List available connectors
saas-scraper list

# Scrape a Slack workspace and stream Documents to stdout (NDJSON)
saas-scraper fetch slack --workspace acme --since 7d

# Save to a file for downstream consumption
saas-scraper fetch notion --workspace acme > docs.ndjson

Programmatic use:

import asyncio
from saas_scraper import BrowserSession, registry

async def main() -> None:
    async with BrowserSession() as session:
        connector = registry.create("slack", session=session, workspace="acme")
        async for doc in connector.discover_and_fetch():
            print(doc.ref.path, len(doc.text or b""))

asyncio.run(main())

Connectors

Connector	Status	Notes
slack	implemented (v0.2)	channel sidebar walk, message pane scrape
github	implemented (v0.3)	file-tree walk + raw fetch, public + SSO repos
gitlab	implemented (v0.3)	gitlab.com or self-hosted via base_url
bitbucket	implemented (v0.3)	bitbucket.org file walk
jira	implemented (v0.3)	Atlassian Cloud issue list + body
confluence	implemented (v0.3)	Atlassian Cloud space page-tree
notion	implemented (v0.3)	sidebar page enumeration + body

All connectors share a single BrowserSession so cookies and SSO state inherit across providers. Virtualised lists (Slack sidebar, Notion sidebar) only see the currently-visible portion in v0.3 — scroll-walking lands in v0.4.

The v0.1.0 release ships the Document protocol, the Chrome session manager, and a working scaffold per connector. Additional providers and per-connector hardening land in subsequent releases — see issues.

Why Chrome and not the API?

• Inherits SSO / MFA / SCIM-locked sessions that don't cleanly expose API tokens to a scanner role.
• Bypasses API quota tiers that throttle org-wide content enumeration.
• Reaches UI-only surfaces (Notion comments, Slack canvas, Jira views).

When an official API exists and is sufficient, prefer that — saas-scraper is the fallback for the cases where it isn't.

Development

uv sync --all-extras
uv run playwright install chromium
uv run pytest
uv run ruff check
uv run mypy src

The default pytest pass exercises plumbing only (Document protocol, registry wiring, CLI helpers). Live browser scrapes against real SaaS providers are not part of CI; run them locally with saas-scraper fetch <connector> --headed so a real Chromium window opens for first-time SSO.

Release

vX.Y.Z tag pushes trigger PyPI trusted publishing via GitHub Actions — no manual token. The first publish requires a one-time Trusted Publisher configuration at https://pypi.org/manage/account/publishing/:

Field	Value
PyPI Project Name	saas-scraper
Owner	plenoai
Repository name	saas-scraper
Workflow name	release.yml
Environment name	pypi

After that, every tag matching v* will publish automatically.

License

AGPL-3.0-or-later.