Know when your MCP server breaks — before your users do.

It means checking a Model Context Protocol server the way a real client uses it: connecting, completing the initialize handshake, listing tools, and optionally calling one — on a schedule. mcptrax runs this full sequence every 1 to 15 minutes and alerts you when any step fails, so problems surface as alerts instead of user reports.

A ping monitor checks whether a URL returns a response. MCP servers can return HTTP 200 while the initialize handshake fails, the tool list is empty, or a tool errors on every call — all invisible to a ping. mcptrax speaks the MCP protocol itself, so it catches the failures agents actually hit.

Tool drift is when your server's tools/list output changes — a tool renamed, removed, or its input schema altered — while the server stays online. It silently breaks every client built against the old contract. mcptrax diffs your tool list on every check and flags drift the moment it appears.

Yes. You can configure a custom header — an Authorization bearer token or an API key header — per monitor, and mcptrax sends it with every check. The value is stored server-side and never exposed in the dashboard again after saving.

Yes. Pro is a monthly or yearly subscription managed through Paddle. Cancel from the billing portal in one click; your plan stays active until the end of the paid period, and your monitors simply revert to the free plan's limits after that.