What problem does it solve?
Turning a local model, Python function, or lab MCP server into a remotely callable tool usually requires public IPs, firewall changes, and manual credential handling. This Skill guides you through wrapping your workload, validating it locally, and sharing it privately through ToolUniverse Platform with browser-based authorization.
Core Features & Use Cases
- Python Function Wrapping: Expose any Python callable as a remote tool using the
@remote_tool decorator and tu serve, with guidance on bounded, JSON-serializable inputs and sanitized errors.
- Existing MCP Endpoint Relay: Forward a loopback Streamable HTTP MCP server through
tuplatform-relay without exposing ports or moving credentials.
- Local Validation Before Sharing: Verify tool discovery, semantic calls, GPU execution, and error behavior locally before enabling private sharing.
- Use Case: A researcher with a GPU-hosted protein scoring model wraps it with
@remote_tool, validates inference locally, then runs tu serve --share so collaborators can call it through ToolUniverse Platform while the model and weights stay on the lab server.
Quick Start
Ask the agent to wrap my Python model function with @remote_tool, validate it locally, and share it privately through ToolUniverse Platform.
