This feature would be life changing. +1

+1

+1, this one really hurts.

My current debug loop on Edge Functions: function errors in prod, I hand it to Claude Code via MCP, it pulls schema, reads migrations, queries tables. Then it hits a wall because it cannot see the logs. So I alt-tab into the dashboard, open Logs, pick the source (Edge Functions / Postgres / API / Auth), set a time range, scroll, copy the stack trace, paste it back into the chat. Every single time.

Everything else in the loop is already programmatic. This one step is the only reason I still keep the dashboard tab open. For intermittent bugs where you want to correlate logs across several invocations it basically kills the workflow, a 30 second fix turns into a 5 minute context switch, and anything you cannot reproduce locally is back to manual spelunking.

What would help:

1. An MCP tool, roughly get_logs(project_ref, source, start_time, end_time, filter), covering edge function, postgres, API, auth logs. Same data the Logs Explorer shows, just callable.
2. A management API endpoint over the Logflare data that we can wrap ourselves.
3. Or just expose the logs as a SQL queryable schema so the existing execute_sql tool can reach them.

The data is already sitting in Logflare and Studio already queries it, there is no new infrastructure needed, only a programmatic surface on top. Even read only, rate limited, last 24 hours would cover most of the AI debugging cases I care about.

Would really appreciate if this got some priority. Right now it is the one big gap between MCP being useful and MCP being the main way I debug.

@mattrossman, @Rodriguespn