Give your coding agent a GPU, and a budget: the Runcrate MCP server in Claude Code, Codex and Cursor A developer released the Runcrate MCP server, a Streamable HTTP endpoint at mcp.runcrate.ai/mcp that lets coding agents in Claude Code, Codex, Cursor and VS Code provision GPUs, run jobs, follow logs and tear machines down. The server confines agents to a single pre-selected environment with a workspace-set monthly budget, and every instance carries max_runtime_minutes (5 minutes to 7 days) and idle_shutdown_minutes (5 minutes to 24 hours) guardrails that no MCP tool can alter. Live pricing on 2026-10-20 listed RTX 4090 from $0.52/hr, A100 80GB from $1.76/hr, H100 from $3.55/hr and H200 from $4.24/hr, with boot time unbilled and prepaid balances. Coding agents are good at writing a training script. They are bad at the part that comes next: finding a GPU, getting a shell on it, running the script, and remembering to shut the machine down. That last step is where the surprise invoices come from. This post shows how to hand that whole loop to the agent you already use Claude Code, Codex, Cursor or any MCP client with the Runcrate MCP server, and how to keep it on a leash while it spends money. The server lives at https://mcp.runcrate.ai/mcp Streamable HTTP . It exposes tools for: Sign-in is OAuth in your browser. There is no API key to paste into a config file. Claude Code claude mcp add --transport http --scope user runcrate https://mcp.runcrate.ai/mcp Then run /mcp inside Claude Code and pick runcrate . Your browser opens to sign in. Codex codex mcp add runcrate --url https://mcp.runcrate.ai/mcp Cursor ~/.cursor/mcp.json , and most other clients: { "mcpServers": { "runcrate": { "url": "https://mcp.runcrate.ai/mcp" } } } VS Code mcp.json in your user profile : { "servers": { "runcrate": { "type": "http", "url": "https://mcp.runcrate.ai/mcp" } } } If you have the Runcrate CLI brew install aeonmindai/tap/runcrate , runcrate mcp install finds the clients on your machine and connects them all. When you sign in, you pick a workspace and one environment . The agent can act only inside that environment, and the environment's monthly budget caps what it can spend. No MCP tool can create or switch environments or change a budget, so the agent cannot raise its own ceiling. Budgets are set in the dashboard by a workspace owner or admin. Two more guardrails sit on every instance the agent creates: max runtime minutes 5 minutes to 7 days : the instance is terminated that long after it starts running, with an alert 10 minutes before. idle shutdown minutes 5 minutes to 24 hours : the instance is terminated once GPU utilization stays under 5% with nobody connected for that long. You do not need to know the parameter names. Say it in the prompt: Launch an H100, stop it after 4 hours or after 30 idle minutes. Check the hardware before you commit What GPUs with at least 48 GB of VRAM are available right now, and what do they cost per hour? The agent calls list gpu types and answers from live prices. Run a training script and come back with the number Run python train.py on an A4000 for at most 2 hours with my volume training-data, then tell me the final loss. What happens under the hood: list volumes finds the id of training-data . run job starts the command with gpu: "A4000" , max runtime minutes: 120 and the volume attached at /workspace . get job logs follows the output until the job is done, and the agent reads the loss from it. get job returns the final status, exit code and cost. The machine stops when the script exits, so there is nothing left running to forget about. Keep a box for interactive work Launch an RTX 4090, install vllm, serve Qwen/Qwen2.5-7B-Instruct on port 8000, and give me an HTTPS link only my workspace can open. Stop it after 3 hours. Here the agent uses create instance with a max runtime, ssh execute to install and start the server, and create instance link to expose port 8000 at an https://8000-