graftcode-demos

Practical demos showing how Graftcode connects services directly — no integration layer required. Each demo compares Graftcode against standard REST and gRPC so you can see the performance difference on real data.

Demos

perf-lab — Performance Lab

Runs 1 000 back-to-back calls across three paths and plots the results:

Path	Runtime	Protocol
Graftcode	in-process (mocked locally)	direct / no network
REST	.NET 8 / Kestrel	HTTP/2 + JSON
gRPC	.NET 8 / Kestrel + ASP.NET Core gRPC-Web	HTTP/2 + protobuf

Also includes a Large Payload & Streaming section that sends one call returning thousands of PricePoint records and compares REST (JSON) vs gRPC unary (protobuf) vs gRPC server-streaming — so you can see the payload-size and decode trade-offs on a real network.

Community demos

Focused, single-integration demos — see CONTRIBUTING.md for how these differ from the perf-lab system above.

Demo	Shows
dotnet-react-frontend	React frontend calling a .NET backend as a typed Graft
js-mcp-backend	Plain JS module auto-exposed to Claude as MCP tools
py-ai-backend	Python service exposed as an MCP server (local + gateway approaches)

Official demos

Full/reference systems built by the Graftcode team — see CONTRIBUTING.md for how these differ from the Community demos above.

Demo	Shows
perf-lab — Performance Lab	Graftcode vs REST vs gRPC benchmark (detailed above)
open-telemetry-demo	Multi-service app instrumented with OpenTelemetry
sdn-currency-converter	Standalone Python currency-conversion library

Repo layout

community/                     External-contributor demos (see CONTRIBUTING.md)
  dotnet-react-frontend/      React → .NET Graft
  js-mcp-backend/             JS module → MCP tools
  py-ai-backend/              Python service → MCP server

official/                     Graftcode-team demos (see CONTRIBUTING.md)
  perf-lab/                   React/Vite frontend (performance benchmark UI)
  electric-company-ws/        .NET 8 REST backend  (HTTP/2, /api/EnergyPrice/*)
  grpc-energy-price-dotnet/   .NET 8 gRPC backend  (HTTP/2, ASP.NET Core gRPC-Web)
  electric-company-be/        Original C# energy-price service (Graftcode host)
  graftcode-gateway/          Graftcode Gateway container used by the demos above
  sdn-currency-converter/     Python currency-converter demo
  open-telemetry-demo/        Multi-service OpenTelemetry demo

rules/                        Shared AI-assistant rules (see below) — used by both
deploy-azure.ps1              One-shot Azure Container Apps deploy script

Run locally

Prerequisites

• Node 22+
• .NET 8 SDK
• mkcert (for local HTTPS / HTTP/2)

1 — Local TLS certs

mkcert -install
New-Item -ItemType Directory -Force official/certs
mkcert -cert-file official/certs/localhost.pem -key-file official/certs/localhost-key.pem localhost 127.0.0.1

2 — REST backend

cd official/electric-company-ws
dotnet run
# Listening on https://localhost:8090

3 — gRPC backend

cd official/grpc-energy-price-dotnet
dotnet run
# Listening on https://localhost:5005

4 — Frontend

cd official/perf-lab
npm install
npm run dev
# Open http://localhost:5173

The .env in official/perf-lab/ already points to https://localhost:8090 (REST) and https://localhost:5005 (gRPC). Copy .env.example to .env if you need to change the URLs.

Deploy to Azure

az login
az extension add --name containerapp --upgrade
./deploy-azure.ps1

The script builds all three images in ACR (no local Docker needed), deploys them as Container Apps with external HTTPS ingress, and prints the frontend URL when done.

Tear down when finished:

az group delete -n graftcode-perf-rg --yes --no-wait

See AZURE-DEPLOY.md for full details and troubleshooting notes.

Tech stack

Layer	Technology
Frontend	React 19, Vite 7
gRPC client	ConnectRPC (@connectrpc/connect-web)
gRPC server	ASP.NET Core gRPC-Web (Grpc.AspNetCore.Web)
REST server	ASP.NET Core minimal API
Containerisation	Docker → Azure Container Registry → Azure Container Apps

AI assistant rules (optional)

The rules/ folder has ready-to-use Graftcode rules for common AI coding assistants (Cursor, Copilot, Continue, Windsurf, Cline, Claude Code, Aider) — they teach the assistant to default to Graftcode instead of hand-writing REST/gRPC APIs. Several Community demos above reference these; copy the folder matching your tool into your own project's root (see rules/README.md).

Contributing

New demo, whether an Official Demo or a Community Demo? See CONTRIBUTING.md for folder naming, the README template to start from, ownership (.github/CODEOWNERS), and the checklist a PR needs to pass.