BastionMCP
Secure broker gateway and stateful trading SaaS that bridges LLM AI agents with brokerage accounts via MCP, encrypting credentials and supporting connectors for Bybit, MetaTrader 5, TradingView, and more.
README
BastionMCP (Sentinel-MCP) — Secure Broker Gateway & Stateful Trading SaaS
BastionMCP is a enterprise-grade Secure Broker Gateway and stateful trading bot system designed to bridge LLM AI Developer Agents (such as Cursor, Cline, and custom LLM scripts) with brokerage accounts and financial exchanges safely, using the Model Context Protocol (MCP).
🌟 The Problem & The Solution
1. The Security Nightmare of Raw Keys
- The Problem: Giving an AI coding agent direct access to raw brokerage API keys, MT5 passwords, or secret tokens is highly risky. A single hallucinated command or unvetted package update could result in disastrous unauthorized trades, account liquidation, or credential leaks.
- The Solution: BastionMCP acts as a security gateway. Sensitive exchange/broker credentials are encrypted in a Credentials Vault using AES-GCM before database storage. LLMs communicate with BastionMCP using client-scoped, revocable API keys, preventing raw credentials from ever being exposed in chat history, logs, or prompt contexts.
2. Fragmentation of Trading APIs
- The Problem: Integrating different exchanges (Bybit), traditional platforms (MetaTrader 5), macroeconomic data sources (FRED), and alert webhooks (TradingView) requires writing custom integration logic for each platform.
- The Solution: BastionMCP standardizes these fragmented protocols under the Model Context Protocol (MCP). Any AI agent supporting MCP can seamlessly discover, query, and interact with these connectors via unified tools.
3. Stateful & Non-Simultaneous Position Management
- The Problem: Standard trading webhooks execute instantaneous "blind" orders. They cannot handle complex, multi-timeframe state transitions or manage existing open positions without active supervision.
- The Solution: BastionMCP features a Stateful Trading Bot that manages open positions. It processes non-simultaneous trading signals (BIAS on 30m/15m -> SETUP on 5m -> ENTRY on 1m) and takes over active trades, executing user-defined preset strategies (Fixed Take Profit, Trailing Stop-Loss, or Contrary Signal Exit) and routing signals to Telegram.
🏗️ System Architecture
The following diagram illustrates the flow of control and data between the AI Agent, the Gateway, database/vault layers, and the execution connectors.
graph TD
subgraph Client Layer
Agent[AI Developer Agent / Cline / Cursor]
Dashboard[Vite Dashboard Frontend]
end
subgraph Security & Routing
Gateway[Cloudflare Worker Gateway]
DB[(Supabase DB)]
Vault[(Secure AES-GCM Vault)]
end
subgraph Execution Connectors
MT5[MetaTrader 5 Bridge]
Bybit[Bybit REST API]
TV[TradingView Webhooks]
FRED[FRED Macro Data]
OHLCV[OHLCV Market Data]
TG[Telegram Command Bot]
end
Agent -->|1. Scoped API Key / MCP Request| Gateway
Gateway -->|2. Validate Key & Fetch Vault| DB
Gateway -->|3. Decrypt Credentials| Vault
Gateway -->|4. Route MCP Request| MT5
Gateway -->|4. Route MCP Request| Bybit
Gateway -->|4. Route MCP Request| FRED
Gateway -->|4. Route MCP Request| OHLCV
Dashboard -->|Manage API Keys, Vaults, & Presets| DB
TV -->|TradingView Alerts| Gateway
TG -->|Command Position Takeover| Bybit
TG -->|Position Alerts & Signals| Agent
📦 Project Workspaces
This repository is organized as a monorepo containing the following components:
| Workspace Path | Description | Tech Stack |
|---|---|---|
[/web](file:///c:/Users/DANIEL/Sentinal-MCP/web) |
Dashboard Frontend: Responsive interface for managing vaults, generating client API keys, configuring trading bot presets, and viewing audit logs. | Vite, Vanilla JS, CSS Glassmorphism |
[/workers/gateway](file:///c:/Users/DANIEL/Sentinal-MCP/workers/gateway) |
Gateway Router: API key validator, rate-limiter, and secure router that forwards requests to connectors. | Cloudflare Workers, TypeScript |
[/workers/connectors/bybit](file:///c:/Users/DANIEL/Sentinal-MCP/workers/connectors/bybit) |
Bybit Connector: MCP server wrapping Bybit futures/spot endpoints for position checking and order routing. | Cloudflare Workers, Bybit API |
[/workers/connectors/mt5](file:///c:/Users/DANIEL/Sentinal-MCP/workers/connectors/mt5) |
MT5 Connector: Bridges commands to MetaTrader 5 terminals (via HTTP/ZeroMQ). | Cloudflare Workers, ZeroMQ/HTTP |
[/workers/connectors/tradingview](file:///c:/Users/DANIEL/Sentinal-MCP/workers/connectors/tradingview) |
TradingView Connector: Handlers for parsing incoming webhook alerts from TradingView charts. | Cloudflare Workers |
[/workers/connectors/fred](file:///c:/Users/DANIEL/Sentinal-MCP/workers/connectors/fred) |
FRED Connector: Macroeconomic indicators tool fetching data from the St. Louis Fed. | Cloudflare Workers |
[/workers/connectors/ohlcv](file:///c:/Users/DANIEL/Sentinal-MCP/workers/connectors/ohlcv) |
Market Data Connector: Live and historical candlestick aggregator. | Cloudflare Workers |
[/workers/telegram-bot](file:///c:/Users/DANIEL/Sentinal-MCP/workers/telegram-bot) |
Telegram Command Bot: Allows users to link Telegram chat IDs, configure Take Profit presets, and manage active position takeovers. | Cloudflare Workers, Telegram Bot API |
[/workers/stripe-webhook](file:///c:/Users/DANIEL/Sentinal-MCP/workers/stripe-webhook) |
Billing Webhook: Handles subscriptions and billing events to automatically adjust user rate limits. | Cloudflare Workers, Stripe |
🚀 Getting Started
1. Local Development
Install monorepo dependencies from the root directory:
npm install
Start the dashboard locally:
npm run dev:web
This starts the Vite dev server for the frontend, which will be accessible at http://localhost:5173.
Run the gateway locally:
npm run dev:gateway
This runs the Gateway Worker using Wrangler on port 8787 (http://localhost:8787).
2. Frontend Hosting (GitHub Pages)
The frontend is a static single-page application that can be hosted entirely for free on GitHub Pages.
This repository is pre-configured with a GitHub Actions workflow (.github/workflows/deploy.yml) that automatically builds and deploys the dashboard.
Setup Instructions:
- Push this repository to your GitHub account:
https://github.com/626gl1ch/sentinal-mcp.git. - On GitHub, navigate to your repository Settings -> Pages.
- Under Build and deployment -> Source, select GitHub Actions (instead of "Deploy from a branch").
- On your next commit (or via a manual trigger in the Actions tab), GitHub will run the build and publish the dashboard to:
https://626gl1ch.github.io/sentinal-mcp/
Once opened, you can bind your Supabase URL, Anon Key, and Gateway Worker URL directly through the Configuration Binding drawer at the bottom right.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.