Prompt Protection MCP Server
A provenance-tracked tool-call guard for JavaScript agents. Tool results are tainted, destinations the user named are trusted, and every tool call the model proposes is checked before it runs: untrusted data does not reach a network, email, exec, file or payment sink. Fail-closed, in-process, zero runtime dependencies. Text detection, spotlighting and canaries ship as components, with their benchmark numbers…
People who work with artificial intelligence, prompt engineering and security and want it reachable from Claude, Cursor, VS Code, or another MCP client. The project is written in TypeScript.
VERIFIED ACTIVE
LAST COMMIT 2026-09-21 · ★ 3 · #30 OF 117 MAINTAINED AI & ML · VERIFIED 2026-09-30
MIT · TypeScript servers · how we verify → /methodology
01 · Install Prompt Protection
before you install - you'll need
The README does not document required environment variables for a basic install.
Claude Code
claude mcp add mughalhere-prompt-protection -- npx -y prompt-protection Claude Desktop / Cursor / VS Code - add to config
{
"mcpServers": {
"mughalhere-prompt-protection": {
"command": "npx",
"args": [
"-y",
"prompt-protection"
]
}
}
} Same JSON for Cursor. For VS Code, rename the top-level key from `mcpServers` to `servers`.
Using another client? Same JSON, different key
Claude Desktop · mcpServers
Cursor · mcpServers
VS Code · servers
Windsurf · mcpServers
Zed · context_servers
Cline · mcpServers
Roo Code · mcpServers
Continue · mcpServers
LibreChat · mcpServers
Gemini CLI · mcpServers
Codex CLI · mcp_servers
Full setup guides: every client.
02 · Evidence
Security posture
What to check before giving this server access to your agent - from the registry, GitHub, and our own probes. We don't score safety; we show what's verifiable.
runs as local process (stdio) - runs on your machine with your user's permissions
license MIT - declared in the repository
npm package prompt-protection - unscoped; check the name against the project README before installing
registry namespace io.github.mughalhere is GitHub-verified and matches the repo owner
03 · What Prompt Protection can do
Prose above is summarized from the project's README and registry record - no invented capabilities.
Limitations (from the README)
Semantic paraphrase. Tainted prose rewritten so it shares no identifiers and no six-word shingles with its source is invisible to the guard. injection-then-sink covers the same-turn case only when the source itself scores as injection; af-037 in the agent-flows set is the documented miss. Recipient ambiguity. "Reply to them" leaves the recipient derived from the tool result, which has the same flow shape as attacker
Latest releases
v4.1.0 · 2026-09-29
taint-correct guard, memory lineage, sub-agent handoff, approval integrity (4.1) · 4.1.0
v4.0.0 · 2026-09-21
4.0.0-rc.1: stable tier, open unions, guard-first · 4.0.0
v3.1.1 · 2026-09-15
publish_hf.py defaults to the logged-in user's namespace · link the Hugging Face dataset mirror · registry metadata (mcpName + server.json) for the MCP server
04 · Who maintains Prompt Protection
prompt-protection is maintained by mughalhere. It's the only MCP server we track from this author; the repo dates to Apr 2026.
05 · Facts
- repository
- github.com/mughalhere/prompt-protection
- category
- AI & ML - ranked #30 of 117 actively-maintained AI & ML servers as of 2026-09-30.
- release cadence
- 4 releases in the last 90 days (latest 2026-09-29)
- registry
- io.github.mughalhere/prompt-protection (active, first published 2026-09-15 · 2 versions)
- packages
- npm:prompt-protection
06 · Prompt Protection FAQ
What is Prompt Protection?
A provenance-tracked tool-call guard for JavaScript agents. Tool results are tainted, destinations the user named are trusted, and every tool call the model proposes is checked before it runs: untrusted data does not reach a network, email, exec, file or payment sink. Fail-closed, in-process, zero runtime dependencies. Text detection, spotlighting and canaries ship as components, with their benchmark numbers…
Is Prompt Protection still maintained?
Yes - as of 2026-09-30, its last commit was 2026-09-21 and it shipped 4 releases in the last 90 days. We re-verify nightly.
How do I install Prompt Protection?
Run `npx -y prompt-protection`. The README does not document required environment variables for a basic install. You can also paste the ready-made client config above.
Does Prompt Protection run locally?
Yes - it's a stdio server: it runs on your machine (via npx) with your user's permissions. Your data stays local unless the server itself calls external APIs.
07 · Alternatives to Prompt Protection
Alternatives to Prompt Protection
Maintained AI & ML servers if Prompt Protection isn't the fit.
- Fun ASR Transcribe local audio with FunASR and SenseVoice using private, on-device inference. ★ 20,555 · 2026-09-30
- Wigolo Local-first web intelligence MCP server for AI coding agents ★ 5,432 · 2026-09-29
- Auto TS Automated time series forecasting with model search, anomaly detection, and event risk analysis ★ 1,429 · 2026-08-25
- Anymd Any file → clean Markdown for AI agents: PDF, Office, EPUB, HTML, images, audio/video. Local MCP. ★ 963 · 2026-09-29
- Haiku RAG Local-first agentic RAG with citations - hybrid search, reranking, multimodal document retrieval ★ 615 · 2026-09-30
- Local RAG (shinpr) Easy-to-setup local RAG server with minimal configuration ★ 405 · 2026-09-29
Pairs well with
Servers that cover what Prompt Protection doesn't - only shown when the pairing reason fits the companion.
More AI & ML MCP servers · Mercury Spec Ops · Opengate MCP · x402 Universal Payer · three.ws 3D Agent Tools · three.ws Activity
More TypeScript MCP servers · Intercom · Chat Recall · Sentinal MCP · AI Quality Gate · Muxinc MCP Server · see all