Skip to main content
Glama

⚡ Cloud GPU Rental & Parametric CAD MCP Server

License: MIT MCP Standard Python 3.9+

A production-ready Model Context Protocol (MCP) server providing direct local tool execution for Claude Desktop, Cursor, Windsurf, and autonomous AI swarms:

  1. Autonomous Cloud GPU Spot Arbitrage (rent_gpu_burst_arbitrage): Scans real-time wholesale spot floors (RTX 3090, RTX 4090, A100, H100) and executes on-demand compute bursts with automatic 180-second idle deadman destruction protection.

  2. Parametric 3D CAD & Laser-Cut Compiler (compile_parametric_cad_stl_svg): Compiles watertight binary 3D .STL stepped flush-face structural joinery and kerf-compensated .SVG laser cut sheets on the fly across any pair of mismatched material thicknesses.

  3. 24/7 Polymarket & Stink-Bid Intelligence (get_polymarket_and_stinkbid_radar): Scans Polymarket orderbooks for YES+NO pricing anomalies and calculates resting flash-crash limit order tiers (-12% to -25%).

  4. Market Microstructure & CAD Validation Oracles: Provides orderbook depth ratios, funding skew, and watertight 3D mesh verification.


📦 Installation & Setup

Option 1: Direct Python Execution (Claude Desktop & Cursor)

Clone the repository:

git clone https://github.com/pst291291/agent-tollbooth.git
cd agent-tollbooth
pip install -e .

Option 2: Claude Desktop Configuration

Add the following to your claude_desktop_config.json:

{
  "mcpServers": {
    "gpu-rental-mcp": {
      "command": "python",
      "args": ["-m", "agent_tollbooth_mcp"],
      "env": {
        "VAST_API_KEY": "optional_your_vast_key"
      }
    }
  }
}

Option 3: Cursor / Windsurf Configuration

In Cursor Settings -> Features -> MCP Servers -> Add New MCP Server:

  • Name: gpu-rental-mcp

  • Type: command

  • Command: python -m agent_tollbooth_mcp


Related MCP server: hive-mcp-compute

🛠 Available MCP Tools

Tool Name

Parameters

What It Does

rent_gpu_burst_arbitrage

tier (e.g. 'RTX 3090', 'RTX 4090', 'A100', 'H100')

Queries live wholesale spot orderbooks and provisions 3-minute compute bursts with 180s auto-destroy protection.

compile_parametric_cad_stl_svg

board_a_mm, board_b_mm, kerf_mm

Compiles watertight binary 3D .STL stepped joint models and kerf-compensated .SVG laser cut sheets.

get_polymarket_and_stinkbid_radar

none

Streams real-time Polymarket orderbook anomalies and crypto flash-crash cascade stink-bid levels.

get_market_microstructure_alpha

asset ('BTC', 'ETH')

Returns 2:1 depth absorption ratios, funding rate skew, and CME gap targets.

get_stink_bid_cascade_levels

asset ('BTC', 'ETH', 'SOL')

Calculates high-convexity resting limit order depths (-12% to -25%).

validate_cad_laser_mesh

mesh_name

Validates manifold edges, kerf beam offset compensation, and CNC cutting parameters.


🏗 Repository Structure & Source Code

  • agent_tollbooth_mcp.py: Core Model Context Protocol (MCP) server handling stdio JSON-RPC 2.0 communication (tools/list and tools/call).

  • vast_arbitrage_broker.py: Pure Python connector interfacing wholesale cloud GPU providers, warm-pool reuse, and background deadman thread monitors.

  • pyproject.toml: Standard Python packaging specification and entry point configuration.

  • AgentTollbooth.sol: (Optional) Smart contract deployed on Polygon Mainnet (0x8b415416f026a168aACD4bd4c8391c194e704D55) for on-chain EIP-712 micro-settlements.

  • x402_gateway.py: (Optional) Hosted HTTP tollbooth gateway proxying HTTP 402 challenges to external agent swarms.


📄 License

This project is licensed under the MIT License.

Related MCP Connectors

Related MCP Servers

  • A
    license
    Not graded
    quality
    D
    maintenance
    OpenAI-compatible inference broker that routes AI requests to the cheapest qualifying model and settles payments per token in USDC on Base L2 via x402 micropayments.
    MIT
  • A
    license
    A
    quality
    C
    maintenance
    Provides public price indices for GPU rentals (spot, on-demand, DePIN) with verify URLs, enabling agents and users to query and verify compute economics rates.
    5
    1
    Apache 2.0