# ClickUp MCP leaks hundreds of orphaned npx/node processes under Claude Code

**TL;DR:** Kill the orphaned processes, then switch to the remote hosted server or pin a fixed version to stop the respawn loop. The local stdio server is being spawned repeatedly without cleanup, consuming gigabytes of memory. Monitor with ps or Activity Monitor after the change to confirm the leak stops.

## The error

```
Hundreds of orphaned npx/node clickup-mcp-server processes accumulating under Claude Code, with heavy memory compression and swap
```

## Fix it

1. List the processes and confirm they are clickup-mcp-server pairs.
   Expected: You see hundreds of them.
2. Kill the orphans.
   Expected: Memory pressure drops.
3. Switch your MCP config to the remote hosted server (clickup-mcp.taazkareem.com/mcp), or pin a patched local version.
   Expected: The config no longer spawns local stdio repeatedly.
4. Restart Claude Code and watch the process count over an hour.
   Expected: It stays flat.

## When this applies

ClickUp MCP under Claude Code spawns ever more npx/node processes that never exit, degrading the machine.

## When this does NOT apply

A single failed startup is a different issue (Node version, token). If the remote server leaks, that is server-side.

## Tool compatibility

taazkareem/clickup-mcp-server 0.14.3, Claude Code on macOS

## Also seen as

- ClickUp MCP memory leak
- clickup-mcp-server processes multiplying
- Claude Code ClickUp CPU memory

## Why it happens

A process-lifecycle bug in the local stdio packaging left each spawned server unreaped when Claude Code recycled the connection, so pairs accumulated. The remote hosted server avoids local process management entirely.

## Edge cases

- macOS shows this as memory compression before swap; check Activity Monitor.
- Killing processes mid-session drops the tool until restart.
- The license key env is still needed for the remote server.