RealAlexandreAI/dsh-cloudflare-browser-run
dsh-cloudflare-browser-run gives the agent real browser access: headless Chrome on Cloudflare's network, so JS-rendered pages, screenshots and PDFs work where the built-in web_fetch (a plain HTTP fetch) comes back empty. The browse tool fetches a public URL into clean markdown (default), or with action=screenshot/pdf returns the saved file path; screenshot and pdf tools save PNG/PDF locally. Sessions are login-capable and WebMCP sites work. Config: cf_api_token (required), cf_account_id (required), optional cf_api_base and output_dir (default OS temp). The plugin identifies itself as a well-behaved bot.
Source repository unavailable. The repository RealAlexandreAI/dsh-cloudflare-browser-run no longer exists on GitHub (HTTP 404 when checked on 2026-10-02). The plugin is still published on npm as dsh-cloudflare-browser-run, which is the install command shown below.
Install
dsh plugin --profile web add dsh-cloudflare-browser-runnpm dsh-cloudflare-browser-run 0.1.4 verified 2026-09-02 (repository field → github.com/RealAlexandreAI/dsh-cloudflare-browser-run; README EN primary). Install per README: dsh plugin --profile web add dsh-cloudflare-browser-run; configure credentials in the profile/settings layer: config cf_api_token (Cloudflare dashboard → API Tokens → 'Browser Rendering: Edit' template) and cf_account_id. It is a port of pi-cloudflare-browser-run built to the dsh (Cordis) plugin spec. Every URL passes an SSRF guard (localhost/private IPs/IPv6 literals/userinfo rejected) before the API is called; the token lives only in the config file.
Compatibility
DSH web profile; Cloudflare Browser Rendering API (cf_api_token with Browser Rendering:Edit + cf_account_id); public web only (SSRF-guarded).
Details
- Repo:
RealAlexandreAI/dsh-cloudflare-browser-run(source unavailable) - Category: Infrastructure & Deployment
- Stars: 1
- Version: npm dsh-cloudflare-browser-run 0.1.4
- Last push: 2026-09-21
- First seen: 2026-08-13
Recent updates
Cloudflare Browser Rendering-backed browse/screenshot/pdf tools; clean markdown from SPA/JS pages; SSRF guard on every URL; login-capable sessions; local output files.
FAQ
- Why not just use web_fetch?
- web_fetch is a plain HTTP fetch — JS pages come back empty. This plugin runs real headless Chrome on Cloudflare's network, so rendered markdown, screenshots and PDFs work.
- What credentials does it need?
- A Cloudflare API token with the 'Browser Rendering: Edit' template and your account id, stored in the plugin config — never logged.
- Can it fetch internal addresses?
- No — every URL passes an SSRF guard (localhost, private IPs, IPv6 literals and userinfo are rejected) before the API is called.