The Ploi MCP tool list-server-logs only accepts a server_id. It always returns the ~25 most recent server logs, including the full output of every script run and the complete automatic-updates output. The Ploi UI (Server > Logs) already has a log type filter (e.g. "[Ploi] System Logs"), a search field and lets you check the output per script run, but the MCP tool exposes neither of those.

Use case: I run a daily, automated check through an AI agent over the MCP. It only needs the output of a few specific scripts and the automatic-updates run of the last 24 hours. Right now every server returns its full set of logs, which is roughly 24k tokens per server (and much more on servers where I ran a verbose script), so most of the cost is output that is thrown away immediately.

Suggested optional parameters for list-server-logs:

• type: same values as the panel filter
• search: text match on the log title (e.g. "Check system status")
• since: ISO 8601 timestamp, return only newer logs
• limit: maximum number of logs
• include_output (boolean, default true): when false, return only title, date, user and status, so an agent can first see what ran and then fetch single outputs

A companion get-server-log tool (by log id) would make this complete.

This would make the MCP much cheaper and faster for monitoring use cases, and it keeps default behaviour unchanged.