Server hangs with no error when SCRAPELESS_KEY is unset (e.g. only SCRAPELESS_API_KEY is set)

[diegopzz]

Found while writing MCP setup docs for editors and agents, against npm 0.4.9 (2026-08-11).

The server reads SCRAPELESS_KEY. Given SCRAPELESS_API_KEY instead — the name used by the Python SDK, the Node SDK examples, and langchain-scrapeless — it starts normally, completes initialize, and then never answers a tools/call. No credential error, no log line, no failure. In an MCP client this looks like the agent freezing mid-task.

Measured, isolated environments

Environment	tools/call scrape_markdown
SCRAPELESS_KEY set	184 chars of markdown returned
only SCRAPELESS_API_KEY set	no response (timed out at 90s)
neither set	no response (timed out at 90s)

tools/list returns all 21 tools in every case, so the server looks healthy right up to the first real call.

Expected

Fail fast with a message naming the variable, and/or accept SCRAPELESS_API_KEY as an alias for consistency with the SDKs. Either would turn a silent hang into a one-line fix for the user.

Minor, same version

serverInfo.version reports 0.2.0 while the package publishes 0.4.9, so clients cannot use it for version gating.