mcpify-cli
v1.0.4
Published
Compile application code into a runnable Model Context Protocol (MCP) server — and auto-register it with Codex, Claude, and VS Code.
Maintainers
Readme
MCPify
Compile application code into a runnable Model Context Protocol server — and auto-register it with your AI clients.
MCPify scans the parts of an app that matter to agents (backend routes & services, frontend actions, OpenAPI specs, Prisma/Drizzle/Mongoose models, webhooks and event listeners), infers multi-step workflows, classifies every action by permission, and generates a complete MCP server. It then writes the server into your local AI clients so the tools show up in the chat bar — no manual config.
Install
npm install -g mcpify-cli
# or run without installing
npx mcpify-cli analyze ./my-appQuick start
mcpify analyze ./my-app
cd ./my-app/.mcpify && npm installRestart your AI client — the generated tools appear automatically.
What you get
Running analyze creates an output directory (./.mcpify by default) containing a
runnable MCP server: server.ts, handlers.ts, tools.ts, workflows.ts,
schemas.ts, AGENTS.md, plus standalone package.json/tsconfig.json.
Every tool is classified as SAFE (agent may call autonomously), REQUIRES_CONFIRMATION (agent must ask first), or BLOCKED (never exposed).
Auto-registration
By default MCPify registers the server into the AI clients it can find:
| Client | Config file |
|---|---|
| Codex | ~/.codex/config.toml |
| Claude Code | <project>/.mcp.json |
| Claude Desktop | platform config dir |
| VS Code | <project>/.vscode/mcp.json |
Existing servers in those files are preserved. Opt out with --no-install, or choose
targets with --clients codex,claude-code.
Commands
| Command | Description |
|---|---|
| analyze [path] | Full pipeline → generate + register the MCP server |
| interactive | Guided setup: pick analyzers and tools |
| frontend [path] | Extract only UI actions (--json for raw output) |
| swagger <file> | Convert an OpenAPI/Swagger spec to MCP tools |
| audit [path] | Static safety audit, no files written |
| simulate [path] | Static audit + (with ANTHROPIC_API_KEY) an AI security battery |
Useful analyze flags
--output <dir>— output directory (default./.mcpify)--swagger <file>/--prisma <file>/--drizzle <path>/--mongoose <path>--no-frontend/--no-events/--no-workflows--ai-enhance— improve tool descriptions via Claude (needsANTHROPIC_API_KEY)--no-install/--clients <list>— control client registration--watch— regenerate on source changes
License
MIT — see the repository.
