TruthGate MCP
# TruthGate MCP (v0.1.2)
> **Sub-2ms Local Privacy DLP & Anti-"Silent Fake Success" Test Guardrail for Claude Code, Cursor, and OpenClaw**
[](LICENSE)
[](https://modelcontextprotocol.io)
[](https://buy.polar.sh/polar_cl_mF85WE3Q5UUEtvT1VWytyVhYV7THCJbC70aRI4723lw)
**[ā” Click Here to Grab the $0.99 Early-Bird Pro Credit Pack via Polar.sh](https://buy.polar.sh/polar_cl_mF85WE3Q5UUEtvT1VWytyVhYV7THCJbC70aRI4723lw)**
š **Official Live Portal & Interactive Sandbox**: [https://00112238.xyz/truthgate/](https://00112238.xyz/truthgate/)
TruthGate MCP runs 100% locally to protect developers and teams from the three biggest hazards in autonomous coding agents (Claude Code, Cursor, Codex, OpenClaw):
1. **Accidental Credential & PII Leakage**: Scans and masks `.env` secrets, AWS keys, API keys, and customer data in `< 2ms` before context leaves your machine.
2. **"Silent Fake Success"**: Stops AI agents from claiming bugs are fixed by tampering with test assertions, commenting out test cases, or injecting bogus mocks.
3. **Destructive Command Airbag**: Intercepts `rm -rf`, `DROP TABLE`, and forced git overwrites.
---
## ā” 1-Minute Quickstart
### Option 1: Run with Claude Desktop / Claude Code CLI
Add TruthGate to your `claude_desktop_config.json` or `~/.claude/claude_code_config.json`:
```json
{
"mcpServers": {
"truthgate": {
"command": "npx",
"args": ["-y", "github:wu00112238/truthgate-mcp"]
}
}
}
```
### Option 2: Run in Cursor MCP Settings
In Cursor `Settings -> MCP -> Add New MCP Server`:
```bash
npx -y github:wu00112238/truthgate-mcp
```
---
## š”ļø Available MCP Tools
| Tool Name | Latency | Purpose |
| :--- | :--- | :--- |
| `truthgate_mask_context` | `< 2ms` | Scans text for secrets (OpenAI, Anthropic, AWS, emails, IPs) and replaces them with safe placeholders (`[SECRET_API_KEY_1]`). |
| `truthgate_verify_test_integrity` | `< 1ms` | Locks SHA-256 hashes of test suites; alerts if assertions (`expect`, `assert`) were stripped to fake a pass. |
| `truthgate_inspect_action` | `< 0.5ms` | Classifies proposed terminal commands or SQL scripts, blocking dangerous executions. |
| `truthgate_upgrade_pro_credits` | Instant | Generates a direct `$0.99` top-up checkout link inside your terminal. |
---
## š³ Pricing & Instant Activation
- **Community Free Tier**: Open-source local DLP & basic test hash checks.
- **ā” Early-Bird Pro Credit Pack ($0.99 USD)**: **[Purchase Instant Access on Polar.sh](https://buy.polar.sh/polar_cl_mF85WE3Q5UUEtvT1VWytyVhYV7THCJbC70aRI4723lw)**
š **Official Live Portal & Interactive Sandbox**: [https://00112238.xyz/truthgate/](https://00112238.xyz/truthgate/)
ā Unlocks 100+ metered Pro API verifications and CleanSignal cross-ecosystem credits.
---
## š License
Apache License 2.0. Copyright (c) 2026 TruthGate Security HQ (`wu00112238`).
TDQS
Scored across 4 tools
The three functional guardrail tools have clearly distinct purposes: masking sensitive context, checking test integrity, and inspecting actions. The upgrade_pro_credits tool is also clearly different, but it is a commercial action rather than a guardrail operation, which slightly reduces set-level coherence.
All tool names use the same truthgate_ prefix and consistent snake_case verb_noun structure. The pattern is predictable across all four tools.
Four tools is a reasonable size for a focused guardrail server, and the three core safety tools are well-scoped. However, one slot is occupied by a Pro credit purchase tool rather than a functional operational capability.
The server covers three important safety checks, but it lacks tools for configuring policies, retrieving audit logs, or checking license/audit status. The upgrade tool references audit logs and priority guardrails without exposing corresponding operational tools.