Commented on Farside Research · Nov 11, 2026
Hey DevHunt, I'm Mike, maker of Farside Research. I built it after one too many AI research answers fell apart the minute someone asked "where's that number from?" Farside searches in rounds, grades what it reads, and writes a synthesis where every claim links to the passage it came from. Gaps and contradictions get their own sections instead of getting papered over. For devs: the same runs work through a REST API and a hosted MCP server, so Codex, Claude, Cursor or any other MCP client can estimate the cost and kick off research on their own. Pipelines re-run a question on a schedule and only ping you when the answer actually moved. A real run, published exactly as it came out: https://farsideresearch.com/sample Trial is 250 credits for 7 days, card required. If you'd plug research into an agent or your own product, what's missing from the API or MCP tools? I'll be here all week.
