WEBGhostingBLOCK
Intelligent stealth browser MCP server for AI agents with 30 tools, 22 anti fingerprint scripts, and LLM powered extraction. Works seamlessly with Cursor, Claude, VS Code, and any OpenAI compatible LLM.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Demo
watch the full demo on YouTube.
What is WEBGhosting
The concept behind WEBGhosting is to tackle the one thing that keeps AI agents trapped in their digital boxes which is the lack of a genuine window to the world. I started with a simple question about whether an agent could navigate a complete browser just like we do, and I wanted to turn that into a reality. It is one thing to have a script that scrapes data, but it is an entirely different beast to build a system that can handle DOM perception and bypass bot detection measures while maintaining the persona of a real human. By focusing on the Model Context Protocol, I created a bridge that lets tools like Claude Code or Cursor step out of the development environment and into the live web, which is a massive leap for autonomous agents.
The way I layered the architecture was very intentional, especially with the Go-based MCP server wrapping Playwright. Building a stealth engine with twenty-two different anti-fingerprint scripts is a level of detail that most people overlook, but it is exactly what makes the difference between a bot that gets blocked immediately and an agent that actually functions. The progression from rigid cinematic demos to a fluid voice-controlled interface highlights how much I learned about where the LLM is most effective. It was fascinating landing on that three-tier execution model because it solves the latency issues that usually kill the vibe of a voice assistant. By using regex for basic tasks and only calling the LLM for the complex stuff, I managed to make the interaction feel instantaneous.
What I built with the Sarvam Voice Agent feels like that Jarvis moment I had been waiting for in tech. It is not just about the code or the thirt
b83348e8c4d2OBSERVED · 2026-10-08Exposed tools (3)
3 read · 0 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
internal-nacl-plugin | read | |
internal-pdf-viewer | read | Portable Document Format |
mhjfbmdgcfjbbpaeojofohoefgiehjai | read |
Trust audit
BLOCKgrade D · trust 66/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | FAIL |
| L2 | Instruction surface (what it tells the agent) | WARN |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (2 observation(s))
- Network
- declared (4 observation(s))
- Shell
- declared (3 observation(s))
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (7)
const obj = eval(objName); // eslint-disable-line no-eval
return [key, eval(`() => ${value}`)()]; // eslint-disable-line no-evalreturn [key, eval(value)]; // eslint-disable-line no-eval
stealth browser
self.recipe_hash = hashlib.md5(recipe_path.encode()).hexdigest()[:10]
print("Launching visually appealing Wavy Voice UI on http://127.0.0.1:8000")webbrowser.open("http://127.0.0.1:8000")Gates applied: no_behavioural_pass.
b83348e8c4d2full audit observations/trust-audit/mcp-server/yranjan06__webghosting.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-08 | b83348e8c4d2 | BLOCK | D | 66 | first audit |
Questions
What is the WEBGhosting MCP server?
Intelligent stealth browser MCP server for AI agents with 30 tools, 22 anti fingerprint scripts, and LLM powered extraction. Works seamlessly with Cursor, Claude, VS Code, and any OpenAI compatible LLM.
What tools does WEBGhosting expose?
3 in total: 3 read-only, 0 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is WEBGhosting safe to connect to an agent?
No — not without reading the findings first. The audit graded it D (66/100) and found 3 critical or high issues in the source. Each one is listed on this page with the file and line it is on.
What credentials does WEBGhosting need?
It reads AI_API_KEY, PROXY_PASSWORD and SARVAM_API_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does WEBGhosting run?
It speaks sse and stdio, so it runs as a local process your client starts.
How current is this page?
The grade is for one exact copy of the source (b83348e8c4d2), read on 2026-10-08. The repository is watched and re-audited when it changes.