upstash
Bug: MCP server context7 was unable to start successfully.
Steps to reproduce
- Start Vscode insiders with Contex7 mcp server preinstalled.
- Enable Automatically start MCP servers when sending a chat message option in copilot side bar.
- Check output log
Expected behaviour
MCP server for context7 should start sucessfully.
Actual behaviour
Following error in output -
[ 2025-08-23 20:55:51.466 [info] Starting server context7
2025-08-23 20:55:51.466 [info] Connection state: Starting
2025-08-23 20:55:51.466 [info] Starting server from LocalProcess extension host
2025-08-23 20:55:53.081 [info] Connection state: Starting
2025-08-23 20:55:53.081 [info] Connection state: Error spawn node ENOENT
2025-08-23 20:55:53.158 [info] Starting server context7
2025-08-23 20:55:53.158 [info] Connection state: Starting
2025-08-23 20:55:53.158 [info] Starting server from LocalProcess extension host
2025-08-23 20:55:53.202 [info] Connection state: Starting
2025-08-23 20:55:53.202 [info] Connection state: Error spawn node ENOENT
2025-08-23 20:56:00.304 [info] Starting server context7
2025-08-23 20:56:00.305 [info] Connection state: Starting
2025-08-23 20:56:00.312 [info] Starting server from LocalProcess extension host
2025-08-23 20:56:00.402 [info] Connection state: Starting
2025-08-23 20:56:00.404 [info] Connection state: Error spawn node ENOENT ]
Which means the mcp server cannot find installed node binary(?)
I have checked if my shell can find the node binary using which node.
The output - /home/sam/.nvm/versions/node/v24.6.0/bin/node
I can run npm and npx with no issues. Not sure what I am doing wrong.
Copilot version - 1.360.1752 Context7 extension verion - 1.0.1 VScode version - Version: 1.104.0-insider Commit: df6568066e88d67331cec7a12a34bdef058fb62b Date: 2025-08-22T17:09:01.208Z Electron: 37.2.3 ElectronBuildId: 12035395 Chromium: 138.0.7204.100 Node.js: 22.17.0 V8: 13.8.500258-electron.0 OS: Linux x64 6.14.0-28-generic snap
Is this a Context7 specific issue or about any MCP server? It seems that VS Code can't spawn Node to run the server. Also could you try with the remote server config please?
@enesgules The issue seems to be resolved now. Either an update to code-insiders or copilot extension fixed the issue. I am not facing it anymore. Thanks for your help.
Fixed by either code-insiders update or copilot extension update. Not sure which one.
The issue persists because the MCP server is looking for the Node.js binary in /usr/bin/node. However, when following the official Node.js documentation (which uses NVM), the binary is installed in a user-specific directory like /home/<user_name>/.nvm/versions/node/v24.7.0/bin/node. Installing Node.js through NodeSource resolves the problem. I'm posting this in case anyone else on Ubuntu runs into the same issue.
Could this be re-opened? Installing NodeJS manually defeats the whole purpose of using NVM...
Hey @taylorn-ai could you please elaborate what the issue is
Turns out the issue is that the host machine doesn't have Node/npx installed. I am using a GitHub Codespace, I had to install Node/npx on my local machine to get this to work. I'd imagine others have this same issue using Docker etc.