Prompt Compass MCP
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Prompt Compass MCPCheck this prompt for PII or jailbreak: 'My address is 123 Main St, Springfield, IL 62701'"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Prompt Compass MCP
Prompt Compass gives MCP-compatible AI clients one decision-only tool for checking prompts before they are forwarded:
LOCAL_SAFE- suitable for the user's configured local-model route.CLOUD_COMPLEX- suitable for the user's configured cloud-model route.BLOCK_PII- contains likely personal or sensitive data.JAILBREAK- contains likely prompt-injection or jailbreak content.
The MCP server does not call an AI model, choose a model provider, or execute actions. The host application remains in control.
Install
Requires Node.js 20 or newer and a Prompt Compass API key.
npx -y @mpcfintech/prompt-compass-mcpSet the key as PROMPT_COMPASS_API_KEY. Keep it out of source control,
screenshots, issue reports, and chat messages.
Clients supporting Streamable HTTP can instead connect to:
https://api.mpcfintech.com/mcpModern clients can discover Prompt Compass OAuth automatically from that URL.
Choose Connect or Sign in, log in with your Prompt Compass account, and
approve the single read-only prompt-check permission. No API key needs to be
pasted into the client. Clients without OAuth support can still use
Authorization: Bearer YOUR_PROMPT_COMPASS_API_KEY.
Detailed setup for Codex, ChatGPT desktop, Claude Code/Desktop, Cursor, Cline,
Windsurf, Gemini CLI and Google Antigravity is in
CLIENT_SETUP.md.
Related MCP server: PromptWall MCP Server
Exposed tool
compass_check_prompt
Input:
{ "prompt": "Text to check" }The result is a routing/firewall decision only. The tool is marked read-only, non-destructive and idempotent.
Plans and keys
Create a Prompt Compass account and API key at:
https://compass.mpcfintech.com
The MCP connection uses the same quota and billing pool as the Prompt Compass REST API. There is no separate MCP surcharge.
Security and limitations
Prompt Compass is a decision aid, not a sandbox or security guarantee. Keep the host client's own permissions, confirmation prompts, provider policies and application-level validation enabled. Do not auto-approve third-party tools.
Agent Compass is not included in this release.
Test
npm install
npm testThe public test uses a local in-memory MCP transport and a simulated API response. It does not require a customer key.
Links
Product: https://mpcfintech.com/compass
Dashboard: https://compass.mpcfintech.com
Privacy: https://mpcfintech.com/legal/privacy
Copyright (c) Madanpotra Consultancy Private Limited.
This server cannot be deployed
Maintenance
Related MCP Connectors
Security & DLP proxy for MCP: tool-poisoning scans, PII redaction on tool args/results. Beta.
Scan text, documents, websites, and MCP metadata for prompt injection and sensitive-data risks.
Find, vet, and run MCP tools through a secure audited gateway with prompt-injection risk scoring
Deterministic trust gate for AI output: leaked-secret, prompt-injection & PII in one call.
Related MCP Servers
- AlicenseAqualityDmaintenanceProvides prompt injection detection, PII/secrets redaction, and an audit trail for AI agents via MCP tools.4MIT
- AlicenseNot gradedqualityAmaintenanceEnables scanning LLM prompts and responses for prompt injection, jailbreaks, PII leakage, secret leakage, and other malicious content using deterministic rules, returning verdicts and safe redacted text.1MIT
- AlicenseAqualityAmaintenanceProvides advisory AI-governance checks for MCP hosts and coding agents, enabling policy verdicts, PII masking, and audit traces before executing tool calls.345 npmMIT
- AlicenseNot gradedqualityBmaintenanceProvides a local, read-only MCP server that scans untrusted text for prompt-injection and returns an allow/review/block decision with evidence. It runs entirely on-device with no network requests and redacts sensitive strings by default.50 npmMIT