convez.ai
MCP Server
sasha-thecornerspore-dev

mcp-command-center

sasha-thecornerspore-dev/mcp-command-center

Ver no GitHub
InicialPluginScore · de 0 a 100

Graphical, AI-assisted command center for connecting AI clients to MCP servers across your system (Windows/macOS/Linux).

Instalação

$npx convez-ai sasha-thecornerspore-dev/mcp-command-center

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
57%
Último commit
há 1 semana
Versão
v0.3.1
Issues abertas
1
Contribuidores
1

README

A tradução para o português deste README ainda está na fila.

🦀 MCP Command Center

One graphical, AI‑assisted control plane for connecting every AI client and every MCP server on your machine.

MCP Command Center detects the MCP‑capable AI tools you already have (Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Continue, Zed), shows them in a single connection matrix, and lets you wire up Model Context Protocol servers with one click — safely writing each client's real config file for you. It also recommends connections with AI, scans your system for opportunities, and keeps an eye out for newly published MCP servers worth adding.

Cross‑platform desktop app (Windows + macOS + Linux), built with Electron + React.


📸 Screenshots

Captured from the running app with representative sample data.

Dashboard — detected clients, live counts, suggestionsConnection Matrix — clients × servers, click to wire
DashboardConnection Matrix
Catalog — searchable server libraryAI Assistant — describe a goal, get a plan
CatalogAI Assistant
Setup — system readiness + guided runtime installProfiles — reusable bundles
SetupProfiles

Regenerate them anytime with npm run screenshots.


✨ What it does

PillarWhat you get
DetectFinds installed MCP clients and reads their current server wiring across OS‑specific config locations.
ConnectA clients × servers matrix. Click a cell to connect/disconnect; review a diff; apply. Every write is backed up first.
CatalogA searchable library of MCP servers from a bundled curated registry, the official MCP registry, and live discovery.
AI AssistantDescribe a goal in plain English (“automate my email + calendar”) → Claude recommends a bundle → you review the diff before anything changes.
ScanInspects your system (git, gh, docker, psql…) and proposes sensible default connections.
TrendsPulls newly published servers from the official MCP registry and surfaces “New & Relevant” cards.
ProfilesSave reusable bundles (e.g. Dev stack = git + github + context7) and apply them across clients in one shot.
Credential identitiesGive a server multiple named credential sets (e.g. user vs. root API keys) with an optional health check, and switch between them with one click; secrets stay OS-encrypted, never in client configs until apply.

Safety first

Every config write is backed up (timestamped), merged (your other settings are preserved), validated, and written atomically, with one‑click restore and automatic rollback on failure. Secrets (API tokens, the Anthropic key) are stored encrypted in your OS keychain via Electron safeStorage — never in plaintext configs.


✅ Requirements

The app is fully self-contained — the installer bundles the Electron runtime, the app, the Anthropic SDK, and the server registry. Nothing else is needed to install and open it.

The MCP servers it wires need a host runtime, because your AI client launches them, not this app: npx-based servers need Node.js, Python servers (git, fetch, time, sqlite) need uv, and container servers need Docker. The Setup tab examines your system, lets you choose a base build (Minimal = Node · Standard = Node + uv · Full = + Docker), and installs any missing runtime via your package manager with one click (or a copy-paste command). See docs/REQUIREMENTS.md for the full breakdown.

🚀 Install

From a release (recommended)

Download the latest installer for your OS from the Releases page:

  • Windows — MCP Command Center Setup x.y.z.exe (installer) or the portable .exe
  • macOS — MCP Command Center-x.y.z.dmg (Apple Silicon + Intel)
  • Linux — MCP Command Center-x.y.z.AppImage

From source

git clone https://github.com/sasha-thecornerspore-dev/mcp-command-center.git
cd mcp-command-center
npm install
npm run dev      # launch in development

Build installers locally:

npm run dist:win    # Windows (nsis + portable)
npm run dist:mac    # macOS (dmg + zip)
npm run dist        # current platform

🧭 Quick start

  1. Launch the app — it auto‑detects your installed clients on the Dashboard.
  2. (Optional) Open Settings and paste your Anthropic API key to enable AI recommendations.
  3. Click Scan system for suggested defaults, or Check for new MCPs to pull the latest from the official registry.
  4. Open the Connection Matrix, click cells to stage connections, then Review & apply.
  5. Save a Profile for any bundle you'll reuse.

🏗️ Architecture

flowchart LR
    subgraph Renderer["Renderer (React)"]
        D[Dashboard] --- M[Connection Matrix]
        M --- C[Catalog]
        C --- A[AI Assistant]
        A --- P[Profiles]
        P --- S[Settings]
    end
    subgraph Main["Main process (Node services)"]
        CD[Client Detector] --> CE[Connection Engine]
        CAT[Server Catalog] --> CE
        SEC["Secrets / safeStorage"] --> CE
        SCAN[System Scanner]
        ADV["AI Advisor (Anthropic)"]
        TW["Trend Watcher (MCP registry)"]
        ST["Preferences / Profiles"]
    end
    Renderer -- "IPC bridge" --> Main
    CE -- "backup, merge, validate, atomic write" --> Files[("Client config files")]
  • Client Detector — locates each client's config and parses current servers.
  • Connection Engine — the only writer; backup → merge → validate → atomic write → restore.
  • Server Catalog — normalizes servers from bundled / official‑registry / web / scanner sources.
  • AI Advisor — turns NL goals into reviewable connection plans (Anthropic).
  • System Scanner / Trend Watcher — suggested defaults and “new & relevant” discovery.

See docs/superpowers/specs for the full design.


🧩 Supported clients

ClientConfig managed
Claude Desktop…/Claude/claude_desktop_config.json
Claude Code~/.claude.json
Cursor~/.cursor/mcp.json
VS Code…/Code/User/mcp.json
Windsurf~/.codeium/windsurf/mcp_config.json
Continue~/.continue/config.json
Zed…/Zed/settings.json

The catalog and client list are data‑driven — adding more is a small change in resources/registry/servers.json and src/main/services/paths.ts.


🛠️ Development

npm run dev         # hot-reloading dev app
npm test            # unit tests (Vitest) — adapters + connection engine
npm run typecheck   # strict TypeScript across main + renderer
npm run build       # production bundle (electron-vite)

Stack: Electron · electron‑vite · React · TypeScript · Tailwind CSS · Vitest · electron‑builder.


🤝 Contributing

PRs welcome — see CONTRIBUTING.md. Good first contributions: add a client adapter, expand the bundled registry, or improve the system scanner probes.

📜 License

MIT © 2026 OpenClaw Community

Ver o original no GitHub

Bom para

Relacionados