Snitch any URL as Markdown. Even Cloudflare/anti-bot protected. Ready for LLM.
Use snitchmd when:
- you need to feed a page to an LLM and raw HTML is too noisy for the context window;
- a plain
curlorfetchreturns an empty shell because the page is rendered by JavaScript; - the site is gated by Cloudflare, reCAPTCHA, or similar anti-bot checks;
- you want one tool that handles all three without picking a scraping engine.
snitchmd https://example.com
That's the whole job. Markdown to stdout, ready to paste into a prompt, a note, or a RAG pipeline.
How it works
Under the hood, snitchmd chains two existing projects so you don't have to: CloakBrowser renders the page past anti-bot checks, and rs-trafilatura strips it down to readable Markdown. It's a small Docker wrapper — no new scraper, no new engine.
Disclaimer: via CloakBrowser, snitchmd looks like a real browser — Cloudflare and other detectors don't flag it as a bot (test results). It can't solve the "click all the traffic lights" kind of CAPTCHAs (reCAPTCHA v2, hCaptcha).
Install
curl -fsSL https://raw.githubusercontent.com/syabro/snitchmd/master/install.sh | bashWithout install:
docker run --platform linux/amd64 --rm -i -v "${XDG_CACHE_HOME:-$HOME/.cache}/snitchmd:/cache" syabro/snitchmd https://example.comShell alias:
alias snitchmd='docker run --platform linux/amd64 --rm -i -v "${XDG_CACHE_HOME:-$HOME/.cache}/snitchmd:/cache" syabro/snitchmd'
Uninstall
curl -fsSL https://raw.githubusercontent.com/syabro/snitchmd/master/install.sh | bash -s -- --uninstallOptional cleanup:
docker rmi syabro/snitchmd
rm -rf "${XDG_CACHE_HOME:-$HOME/.cache}/snitchmd"Run
snitchmd https://example.com
Save the Markdown:
snitchmd https://example.com > page.mdBenchmarks
Raw HTML vs snitchmd output, measured in tokens (tiktoken cl100k_base).
| URL | curl | snitchmd | reduction |
|---|---|---|---|
| https://www.cloudflare.com/learning/bots/what-is-a-bot/ | ❌ HTTP 403 | 0.8k | — |
| https://docs.docker.com/engine/install/ | 187.0k | 0.9k | 100% |
| https://en.wikipedia.org/wiki/LLM | 222.7k | 29.7k | 87% |
| https://github.com/anthropics/anthropic-sdk-python | 127.0k | 0.2k | 100% |
| https://corrode.dev/blog/bugs-rust-wont-catch/ | 22.0k | 4.7k | 79% |
Measured 2026-04-29.
Cache
Each successful fetch is cached on disk by URL + relevant flags. Re-running the same command reads from cache (no browser launch, no extraction).
- Location:
${XDG_CACHE_HOME:-$HOME/.cache}/snitchmd/on the host (mounted into the container as/cache). - Bypass and refresh a single URL:
snitchmd --no-cache https://example.com. - Wipe everything:
rm -rf ~/.cache/snitchmd. - Output-only flags (
--json,--html-output) don't affect the cache key, so the same URL is cached once across them.
Update
docker pull syabro/snitchmd
Troubleshooting
When the tool itself or its install is broken. Content-shape problems (empty output, missing sections, too much chrome) are pipeline-tuning decisions — see snitchmd --help.
| Symptom | Fix |
|---|---|
snitchmd: command not found |
Re-run the install one-liner; ensure ~/.local/bin is on PATH |
Docker is not installed or not in PATH |
Install Docker — https://docs.docker.com/get-docker/ |
Docker is installed, but the Docker daemon is not running |
Start Docker (Docker Desktop on Mac/Windows; the Docker service on Linux) |
permission denied on the Docker socket |
Add your user to the docker group, then re-login (Linux) |
| Image pull fails on first run | Check connectivity; retry with docker pull syabro/snitchmd |
--platform linux/amd64 warning on Apple Silicon |
Expected — runs under Rosetta emulation (slower but works) |
| Cache directory not writable | Fix ownership of ${XDG_CACHE_HOME:-$HOME/.cache}/snitchmd, or delete it to let it recreate |
Exit codes. 0 success, 1 runtime error (browser, network — full message on stderr), 2 extraction returned empty content (page was a loading shell, hit a wall, or has no main content).
JSON output
Use --json when you need metadata such as the final URL, page title, extraction quality, and Markdown length:
snitchmd https://example.com --json
All options
snitchmd --help
Roadmap
Open ideas live in tasks.md in mdtask format. Browse with pnpx mdtask list (or any markdown viewer).
Development
docker build -t snitchmd:local .
docker run --rm snitchmd:local https://example.comLicense
MIT