Jungle Grid MCP Server

Run Jungle Grid GPU workloads from MCP-aware AI hosts such as Claude Desktop,Cursor, Windsurf, and MCP Inspector.

The server runs locally over stdio and forwards tool calls to the Jungle GridREST API with your API key.

Requirements

Node.js 18 or newer
A Jungle Grid API key
Optional: JUNGLE_GRID_API_URL for a self-hosted orchestrator

For the full submit workflow, the API key needs jobs:write. That scope allowsestimate, submit, polling, cancellation, and logs for jobs owned by the key'saccount. list_jobs still requires jobs:read.

Quick Start

JUNGLE_GRID_API_KEY=jg_... npx -y @jungle-grid/mcp

On Windows PowerShell:

$env:JUNGLE_GRID_API_KEY = "jg_..."
npx -y @jungle-grid/mcp

The server uses stdio, so a successful manual launch appears to wait for MCPmessages. If JUNGLE_GRID_API_KEY is missing, it exits with a clear error.

Claude Desktop

Add this to claude_desktop_config.json, then fully restart Claude Desktop.

{
  "mcpServers": {
    "junglegrid": {
      "command": "npx",
      "args": ["-y", "@jungle-grid/mcp"],
      "env": {
        "JUNGLE_GRID_API_KEY": "jg_..."
      }
    }
  }
}

Windows config path:

%APPDATA%\Claude\claude_desktop_config.json

macOS config path:

~/Library/Application Support/Claude/claude_desktop_config.json

Cursor or Project MCP Config

For a checked-in project config, avoid committing secrets. Put the API key inthe environment used to launch Cursor and keep the config secret-free.

{
  "mcpServers": {
    "junglegrid": {
      "command": "npx",
      "args": ["-y", "@jungle-grid/mcp"]
    }
  }
}

For a local, uncommitted config, you can include the key directly:

{
  "mcpServers": {
    "junglegrid": {
      "command": "npx",
      "args": ["-y", "@jungle-grid/mcp"],
      "env": {
        "JUNGLE_GRID_API_KEY": "jg_..."
      }
    }
  }
}

Self-Hosted Orchestrator

JUNGLE_GRID_API_URL defaults tohttps://api.junglegrid.dev. Override it when your host shouldcall a different orchestrator.

{
  "mcpServers": {
    "junglegrid": {
      "command": "npx",
      "args": ["-y", "@jungle-grid/mcp"],
      "env": {
        "JUNGLE_GRID_API_KEY": "jg_...",
        "JUNGLE_GRID_API_URL": "https://your-orchestrator.example.com"
      }
    }
  }
}

Tools

estimate_job: estimate GPU tier, region, duration, and credit cost.
submit_job: submit an asynchronous GPU workload with optional environment values.
get_job: fetch current job status and details.
list_jobs: list recent jobs for the authenticated account.
cancel_job: cancel a pending, queued, or running job.
get_job_logs: fetch stdout, stderr, and exit information.
stream_job_logs: stream live logs until completion or timeout.
list_job_artifacts: list managed artifacts uploaded for a job.
get_artifact_download_url: create a signed download URL for one managed artifact.

Real-Time Job Pattern

Use submit_job to start work, stream_job_logs for live output, thenlist_job_artifacts after completion to retrieve saved files.

{
  "command": ["python", "-c", "import os; exec(os.environ['CODE'])"],
  "environment": {
    "CODE": "import os, json\nos.makedirs('/workspace/artifacts', exist_ok=True)\nwith open('/workspace/artifacts/output.json','w') as f:\n    json.dump({'status':'ok'}, f)"
  }
}

This is the recommended pattern when the real Python payload is too long to fitcomfortably in the command array.

For managed jobs, Jungle Grid automatically creates /workspace/artifacts anduploads any regular files written there. Users do not need to create signedupload URLs or call artifact completion endpoints manually.

Local Development

npm install
npm run build
JUNGLE_GRID_API_KEY=jg_... node dist/index.js

Inspect the server with MCP Inspector:

JUNGLE_GRID_API_KEY=jg_... npx @modelcontextprotocol/inspector node dist/index.js

Publishing

Verify the package before publishing:

npm run build
npm pack --dry-run

Publish the scoped package publicly:

npm publish --access public

Troubleshooting

JUNGLE_GRID_API_KEY environment variable is required: add the key to thehost config env block or to the environment that launches the host.
Tools do not appear: fully quit and reopen the MCP host after editing config.
Old package version: pin a version in config, for example["@jungle-grid/[email protected]"], or clear the npx cache.
API calls fail: confirm the key is valid and JUNGLE_GRID_API_URL points tothe orchestrator you intend to use.

Contributors wanted

We are opening up the Jungle Grid MCP server for contributors interested in AI agents, MCP, developer tools, and workload execution.

Good first areas:

Improve docs
Add example prompts
Add tests for MCP tool handlers
Add Docker support
Improve GitHub Actions
Build integration examples

Start with issues labeled good first issue.

Jungle Grid MCP Server