Portable MCP runtime tutorial

This tutorial is the repo reference application. It composes discoverable capabilities into workflows, lets AI propose without becoming authoritative, and explains the execution through CLI, JSON, MCP, and a browser app.

By Enrico Piovesan Updated June 22, 2026

Tutorial route

Keep the final chapter’s navigation close to the lesson so the source, the examples index, and the previous step remain visible up front.

Open source folder All UMA examples Previous: Discoverable Decisions Live reference application

What you will learn

how WASM MCP exposes discovery and invocation surfaces
how the UMA runtime validates and executes selected capabilities
how AI-backed capabilities participate with explicit fallback reporting

Prerequisites

Rust 1.76 or newer
wasm32-wasip1 target
Node.js 20 or newer
npm
Wasmtime on your PATH

Run this setup command before the lab if your machine does not already have the target installed.

rustup target add wasm32-wasip1

Full tutorial

Enter the example
```
cd chapter-13-portable-mcp-runtime
```
Set up pinned model artifacts
```
./scripts/setup_models.sh
```
Build the PlannerAI WASI module
```
./scripts/build_planner_ai_wasi.sh
```
Build the SummarizerAI WASI module
```
./scripts/build_summarizer_ai_wasi.sh
```
Build the TranslatorFr WASI module
```
./scripts/build_translator_ai_wasi.sh
```
List workflows
```
./scripts/list_labs.sh
```

Run the French AI report workflow

./scripts/run_lab.sh use-case-2-ai-report

Render the workflow as JSON

cargo run --manifest-path rust/Cargo.toml -- render use-case-2-ai-report json

Run the MCP server smoke check
```
./scripts/smoke_mcp_server.sh
```
Run the full chapter smoke path
```
./scripts/smoke_portable_mcp_labs.sh
```

What to inspect

After each command, look for these proof points. They are the signals that connect the code example back to the UMA architecture claim.

capability selection
runtime validation result
structured report JSON
explicit provider fallback reporting when used

Acceptance check

The chapter-level validation path is:

./scripts/smoke_portable_mcp_labs.sh

Return to the repository root for the final acceptance gate:

cd ..
./scripts/smoke_reader_paths.sh

Where to go next

This chapter closes the validated learning path and links to the live reference application.

Open source folder All UMA examples Read Chapter 13: Agents and MCP → Previous: Discoverable Decisions Live reference application