Everything reads and nothing writes: no voting, no commenting, no submitting, so the token cannot do damage even in an agent loop that goes wrong. The pagination model is visible in every signature — counts default to 10 and cap at 20, with an after cursor for the next page — which means a broad question becomes several calls, and each response carries the current rate limits so the ceiling is visible before it is hit. Comments arrive attached to their post rather than as a standalone feed, so reading a discussion starts from the launch it belongs to.
A read-only server over the Product Hunt API — posts, comments, collections, topics and users, with rate-limit information returned alongside every result.
- Posts filtered by topic, featured status, posted-before and posted-after dates or URL, ordered by RANKING, NEWEST, VOTES or FEATURED_AT, up to 20 per call
- One post in depth, returned with its comments, and a separate comment listing ordered by NEWEST, OLDEST or VOTES for reading a launch's reception
- Collections by ID or slug, or listed by the user who made them, by a post they contain, ordered by follower count or newest
- Topics looked up by slug and searched by name, including the topics a particular user follows
- People: a user by ID or username with the posts they MADE or VOTED for, and the account the token belongs to
- check_server_status for authentication state and current rate limits before a long run of calls
A Product Hunt API token for the server to authenticate with — check_server_status reports whether it is valid, and get_viewer fails on a token that has expired.
One command plus a key — pip install product-hunt-mcp, then supply credentials