Install, configure, and update MCP servers across all your AI clients in one command.
npx mcp-fleet install githubmcpm is a CLI that manages Model Context Protocol servers — the tools that give AI assistants like Claude, Cursor, or Windsurf access to your files, APIs, databases, and more.
Instead of editing JSON config files by hand for each client, mcpm handles everything: it detects your installed AI clients, writes the right config format for each one, and keeps your servers up to date.
# Install a server into all detected AI clients
npx mcp-fleet install github
# Search the registry (1000+ servers)
npx mcp-fleet search database
# Install a curated bundle
npx mcp-fleet install @bundle/webdev
# Check what's installed and healthy
npx mcp-fleet doctorOr install globally:
npm install -g mcp-fleet
mcpm install github| Client | Config written automatically |
|---|---|
| Claude Code | ~/.claude.json |
| Cursor | ~/.cursor/mcp.json |
| VS Code Copilot | ~/.vscode/mcp.json |
| Windsurf | ~/.codeium/windsurf/mcp_config.json |
| Zed | ~/.config/zed/settings.json |
| Gemini CLI | ~/.gemini/settings.json |
| Cline | VS Code extension storage |
| Continue | ~/.continue/config.json |
| OpenAI Codex | ~/.codex/config.json |
| Runtime | Command | Example |
|---|---|---|
| Node.js | npx |
npx -y @modelcontextprotocol/server-github |
| Python | uvx |
uvx mcp-server-fetch |
| Docker | docker run |
docker run -i --rm mcp/puppeteer |
| Go | go run |
go run github.com/org/server@latest |
| Deno | deno run |
deno run --allow-net jsr:@scope/server |
mcpm install <server|@bundle/name> Install one or more servers
mcpm uninstall <server> Remove a server from all clients
mcpm update Update all installed servers
mcpm search [query] Browse the registry
mcpm info <server> Show server details
mcpm list List installed servers
mcpm outdated Check for outdated packages
mcpm doctor Diagnose broken servers
mcpm sync Sync from .mcpmrc project file
mcpm rollback Restore the latest config snapshot
mcpm export [file] Export your config
mcpm import <file> Import a config
mcpm publish Submit a server to the registry
mcpm run <server> Run a server directly (stdio)
Pin your project's MCP servers in a .mcpmrc file and let teammates sync in one command:
{
"servers": ["github", "postgres", "filesystem"],
"bundles": ["@bundle/webdev"]
}mcpm sync
mcpm sync --dry-run --receipt .mcpm/sync-receipt.json
mcpm rollbackBefore writing configs, mcpm stores rollback snapshots in ~/.cache/mcp-fleet/rollback/ because Claude Code, Cursor, and other client config files are global/user-level, not project-local. Keep rollback snapshots out of git because they may contain local config values.
For team review flows, see rendered output drift receipts: a safe dry-run artifact for checking which client config files would change before write/rollback.
The registry contains 1000+ verified MCP servers across 5 runtimes.
Browse at azerdsq131.github.io/mcpm or search with:
mcpm search <query>
mcpm search --bundles| Bundle | Description |
|---|---|
@bundle/webdev |
Web development tools |
@bundle/ai |
AI & LLM integrations |
@bundle/data |
Data & databases |
@bundle/devops |
DevOps & infrastructure |
@bundle/productivity |
Productivity & scheduling |
@bundle/cloud |
Cloud provider APIs |
@bundle/finance |
Finance & trading |
@bundle/social |
Social media & comms |
@bundle/media |
Media & content |
@bundle/startup |
Startup essentials |
@bundle/ai-tools |
AI tooling & agents |
mcpm publishThe CLI will guide you through the details and open a pull request automatically.
MIT



