You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When Sandbox.create() is called with an mcp config and the MCP gateway fails to start (non-zero exit code or exception from commands.run), the already-created sandbox is never killed. This leaves orphaned sandboxes consuming resources until they time out.
Affected: Python SDK (async + sync) and JS/TS SDK.
Steps to reproduce:
Call Sandbox.create(mcp={"invalid_server": {}}) with a config that causes mcp-gateway to fail
Observe the exception is raised
The sandbox remains running (visible via Sandbox.list())
Expected: The sandbox should be killed before the error is propagated.
Activity
fix(sdk): clean up sandbox when MCP gateway startup fails (#1547)
linear-code commentedon Jul 23, 2026
SDK-276
fix(sdk): clean up sandbox when MCP gateway startup fails (#1548)
mishushakov commentedon Aug 10, 2026
Hey, this has been fixed in :
• JS SDK (e2b) v2.38.1
• Python SDK (e2b) v2.37.1
Thanks for your patience!