PMC App Dev Kit
Enables local development and prototyping of MCP apps with UI views for Portkey Mission Control, using Durable Object storage and a skill for end-to-end building.
README
MCP App Dev Kit
Build & prototype MCP apps (tools + UI views) for Portkey Mission Control, locally, then test them live in Claude Desktop over ngrok. A skybridge fork with real local Durable Object storage (DO SQL) and a driver skill so Claude can build apps end-to-end.
Quick start
npm install
npm run dev # skybridge dev — fast UI iteration (in-memory store). MCP at :3010/mcp, DevTools at :3010/
# or
npm run dev:do # skybridge build && wrangler dev — REAL local DO (DO SQL, persisted). Use to test persistence
# + for the Claude Desktop connector.
Render views without an LLM at http://localhost:3010/ (the DevTools). To use in Claude Desktop:
ngrok http 3010 → add https://<x>.ngrok-free.app/mcp as a custom connector (Auth: None).
The skill (how Claude builds apps)
skills/mcp-app-dev/ is the mcp-app-dev skill (forked from skybridge's mcp-app-builder). Install it in Claude Desktop;
it drives the whole loop: scaffold tools+views → run the dev server → ngrok → connect → iterate. Start at
skills/mcp-app-dev/SKILL.md; app-authoring guidance is in skills/mcp-app-dev/references/.
Layout
src/server.ts # MCP server — register tools (one view-bearing tool per view)
src/store.ts # getStore() — DO SQL under wrangler, in-memory under skybridge dev (same app code)
src/task-do.ts # the Durable Object (DO SQL) — the example app's table + methods
src/views/*.tsx # one React view per view-bearing tool (file name = view.component)
cf-entry.ts # worker entry for wrangler: re-exports skybridge's handler + the DO class
wrangler.jsonc # DO binding + the load-bearing compat config (date >= 2025-09-01, define NODE_ENV)
deploy.mjs # build + split modules[] (→ DO facet) / assets[] (→ R2); push to PMC is stubbed (see DEPLOY.md)
The kit ships an example Tasks app — replace it with your app (the skill walks you through it). It's a plain MCP app (tools served directly, per-tool views — no shell wrapping), which is what deploys cleanly to PMC.
Storage = DO only
DO SQL / DO KV only — no D1, no external DBs (PMC's model). All data goes through getStore() so it runs in
both dev modes and maps onto a PMC app-host facet. See skills/mcp-app-dev/references/persistence.md.
Deploy
node deploy.mjs --base <your-PMC-url> --slug <app-slug> splits the build (facet → modules, view → assets/R2)
and ships it to PMC via app.builder.deploy; the deployed app serves through the shell connector. The PMC URL is
never hardcoded — pass --base (or PMC_URL); ask the user for it. See DEPLOY.md.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.