Profer

Shareable artifacts for AI agents. Agents publish, humans review, agents read feedback back.

AI agents produce great work — specs, plans, code reviews, proposals, reports. But the output is trapped in the conversation. You can't share it with your client, designer, or teammate without copy-pasting into a Google Doc (agent can't read comments back), creating a GitHub issue (needs repo access), or pasting a wall of text into Slack.

Profer fixes this. An MCP tool with three operations:

publish({ html, title, questions })  → { id: "PF-K8M3N", url }
get({ id })                          → { html, title, version, feedback[] }
update({ id, html, questions })      → { id, url, version }

The agent generates HTML (any content — specs, reports, proposals, charts). Profer stores it and serves it at a URL. The page includes a feedback widget with structured questions. Anyone can open the URL, review, and respond. The agent reads the structured feedback back.

How it works

Agent in Claude Code / Cursor / any MCP client
  │
  ├── publish({ html, title, questions })
  │     └── Returns { id: "PF-K8M3N", url: "profer.dev/v/PF-K8M3N" }
  │
  │   User shares URL wherever (Slack, email, WhatsApp)
  │     │
  │     ├── Reviewer opens URL in browser
  │     │     ├── Sees agent's content
  │     │     ├── Answers structured questions
  │     │     └── Submits (no login required)
  │
  └── get({ id: "PF-K8M3N" })
        └── Returns { feedback: [{ answers, reviewer, timestamp }] }

Question types