Sentinelgate

Name: Sentinelgate
Author: Sentinel-Gate

@Sentinel-Gate

Sentinelgate について

Open-source MCP proxy for AI agent access control. Intercepts every tool call with CEL policies, RBAC, full audit trail, content scanning, and Admin UI. Single binary, zero dependencies.

#security #access-control #proxy #audit #rbac #mcp

基本情報

カテゴリ

その他

トランスポート

stdio

公開者

Sentinel-Gate

投稿者

SentinelGate

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "sentinelgate": {
      "type": "http",
      "url": "http://localhost:8080/mcp",
      "headers": {
        "Authorization": "Bearer <your-api-key>"
      }
    }
  }
}

ツール

ツールは検出されませんでした

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is SentinelGate?

SentinelGate is an MCP (Model Context Protocol) proxy that intercepts every tool call, shell command, and file read before it executes. It enforces deterministic rules and provides a full audit trail. Designed for developers and security teams running AI agents with MCP.

How to use SentinelGate?

Install with one command (macOS/Linux: curl ... | sh; Windows: irm ... | iex) or download a release binary. Start with sentinel-gate start, then point any MCP-compatible client to http://localhost:8080/mcp with an API key. Policies are managed via the Admin UI at http://localhost:8080/admin.

Key features of SentinelGate

Single binary, zero dependencies, sub‑millisecond overhead
Deterministic enforcement with CEL‑powered rules
MCP‑native proxy that aggregates multiple upstream servers
Full audit trail with identity, decision, and timestamps
Admin UI for policy editing, testing, and monitoring
Identity and access control with API keys and roles
Content scanning for PII, secrets, and IPI
Session‑aware policies and behavioral drift detection
Built‑in red team testing with 30 attack patterns

Use cases of SentinelGate

Block prompt injection and hallucination‑driven attacks
Enforce least‑privilege access per agent and per tool
Audit every AI agent action for compliance
Run red team simulations to test security posture
Isolate agent workloads without modifying agent code

FAQ from SentinelGate

How does SentinelGate differ from AI‑based security solutions?

SentinelGate uses deterministic rules (CEL), not AI judgment. deny delete_* always means denied.

What are the runtime dependencies?

None. SentinelGate is a single binary with zero dependencies and sub‑millisecond overhead.

What data does SentinelGate store?

It stores audit logs, policies, session recordings, and configuration. All persisted automatically via the Admin UI. Optionally exportable.

Is SentinelGate a full OS sandbox?

No. It is an MCP proxy that controls tool access through the MCP protocol. For full OS‑level isolation, combine with containers or VMs.

How do agents authenticate?

Agents authenticate with an API key passed via the Authorization: Bearer <key> header. API keys are managed in the Admin UI with per‑identity roles.