MCP.so
ログイン
B

Brackoracle

@brack-6

リポジトリ

Brackoracle について

Runtime security oracle for AI agents — scan prompts, tool calls, and outputs before execution. Pay-per-call via x402/USDC, no accounts needed.

#security#ai-agents#prompt-injection#safety#llm#x402#micropayments#runtime#guardrails#multi-agent

基本情報

カテゴリ

開発者ツール

トランスポート

stdio

公開者

brack-6

投稿者

Baker

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": {
    "brackoracle": {
      "command": "python3",
      "args": [
        "/path/to/brackoracle_mcp.py"
      ],
      "env": {
        "BRACKORACLE_URL": "https://brack-hive.tail4f568d.ts.net",
        "BRACKORACLE_KEY": "AGENTFAST"
      }
    }
  }
}

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 BrackOracle?

BrackOracle is a runtime security oracle for AI agents that scans prompts, tool calls, and outputs before execution. Every decision is HMAC-SHA256 signed and written to an append-only audit log for cryptographic verification. Designed for autonomous agents and compatible with OpenClaw.

How to use BrackOracle?

Call POST /v2/scan with a JSON body containing the content to scan, optionally including the X-Free-Tier: AGENTFAST header for free tier access. Use the Python SDK via BrackClient for programmatic integration. Paid tier uses x402 micropayments via USDC on Base—no accounts, API keys, or subscriptions needed.

Key features of BrackOracle

  • HMAC-SHA256 signed, append-only audit log
  • Hash-only logging; raw prompts are never stored
  • 13+ specialist endpoints for specific detection tasks
  • Zoa cognitive blindspot audit with four independent faculties
  • Prompt lineage tracking for multi-agent contamination
  • Self-hostable under Apache 2.0 license

Use cases of BrackOracle

  • Detect prompt injections and jailbreak attempts before LLM reasoning
  • Validate tool calls against scope escalation and trust laundering
  • Scan agent outputs for secret leakage before publishing
  • Audit cognitive blindspots in agent reasoning with Zoa
  • Track injection contamination across multi-agent pipelines

FAQ from BrackOracle

What is the free tier?

The first 200 calls per IP are free when using the X-Free-Tier: AGENTFAST header—no account required.

Are raw prompts stored anywhere?

No. BrackOracle uses hash-only logging throughout the system; raw prompt content is never persisted.

How can I verify a decision cryptographically?

Call GET /audit/verify/:id to retrieve and verify the HMAC signature for any request row.

Can I self-host BrackOracle?

Yes, the Apache 2.0–licensed codebase runs on a single N100 mini PC with Node.js and Python FastAPI services.

How does payment work?

Paid usage uses x402 micropayments in USDC on Base—agents pay autonomously without accounts or API keys.

コメント

「開発者ツール」の他のコンテンツ

MCP Unity Editor (Game Engine)

CoderGamester

Model Context Protocol (MCP) plugin to connect with Unity Editor — designed for Cursor, Claude Code, Codex, Windsurf and other IDEs

1.8K

Huoshan Test

volcengine

2.4K

Stakpak Agent CLI

stakpak

Ship your code, on autopilot. An open source agent that lives on your machines 24/7 and keeps your apps running. 🦀

1.6K

Unity MCP (Server + Plugin)

IvanMurzak

AI Skills, MCP Tools, and CLI for Unity Engine. Full AI develop and test loop. Use cli for quick setup. Efficient token usage, advanced tools. Any C# method may be turned into a tool by a single line. Works with Claude Code, Gemini, Copilot, Cursor and any other absolutely for fr

3.5K

mcp-excalidraw

yctimlin

MCP server and Claude Code skill for Excalidraw — programmatic canvas toolkit to create, edit, and export diagrams via AI agents with real-time canvas sync.

2.1K