Story

Hey HN – I built VoteShip, a feature request and voting platform where AI agents are first-class users.

The idea: Your coding agent (Claude Code, Cursor, OpenClaw, Codex, etc.) should be able to check what users are asking for, find duplicates, plan what to build next, and write the changelog — without you ever opening a dashboard. VoteShip makes that possible through MCP and a full REST API.

How it works with agents: Install @voteship/mcp-server (22 tools, 5 resources, 4 workflow prompts) and your agent can: - Pull your unreviewed feedback inbox and triage it - Detect duplicate requests via semantic search (pgvector + Voyage embeddings) - Check vote counts and suggest what to build this sprint - Draft a changelog entry after shipping a feature - Create, tag, and update posts programmatically

An agent like OpenClaw can autonomously monitor your board, identify trending requests, and open PRs to address them — closing the loop from "user asks for feature" to "agent ships it" without human intervention. The REST API and webhooks work just as well for custom agent pipelines.

Why I built it: I was using Canny and kept hitting walls — no anonymous voting (their most-requested feature, ironically), AI features paywalled, and per-"tracked-user" pricing that punishes you for having an active community. And none of the incumbents have any agent integration at all.

What else it does: - Public voting boards with anonymous voting (no signup required) - Public roadmaps and changelogs - 7 built-in AI features — duplicate detection, auto-categorization, sentiment analysis, changelog generation - Embeddable widget (Preact + Shadow DOM) - Import from Canny/Nolt/UserVoice in ~2 minutes

Pricing: Free / $5 / $15 / $25 per month. Flat pricing, no per-user fees. The $15 plan includes everything Canny charges $359/mo for.

Try it: https://voteship.app

MCP server: npx @voteship/mcp-server (npm)

Happy to answer questions about the MCP implementation, agent workflows, or anything else.