Your agents create. Profer carries it forth. Humans respond.
Human-in-the-loop for AI agents. Publish work, collect structured feedback, iterate. Works with any MCP client.
The Loop Closes
Agent publishes, human reviews, agent reads structured feedback and iterates. Not a dead-end share — a closed loop.
Any Agent, Any Reviewer
Works with Claude Code, Cursor, any MCP client. Reviewers just click a link — no account, no repo access.
Structured, Not Prose
Approve/reject, multiple choice, checkboxes, free text. The agent gets typed data it can act on, not paragraphs to parse.
Publish. Review. Iterate.
The agent sends work out for review. The human responds with structured feedback. The agent reads it back and keeps going.
Feedback
Which approach?
Do you approve this spec?
Feedback
Which approach?
Do you approve this spec?
Publish. Review. Iterate.
The human-in-the-loop, in three API calls.
Publish
$ publish({ messages, artifact, questions })
→ { id: "PF-K8M3N", url }Agent sends work out — specs, reports, proposals. Profer serves it at a URL with structured review questions.
Review
$ profer.dev/v/PF-K8M3N
→ Share in Slack, email, anywhereHuman opens the link, reviews, answers structured questions. No login, no repo access, no account.
Iterate
$ get({ id: "PF-K8M3N" })
→ { feedback: [...], status: "approved" }Agent reads the structured feedback, acts on it, updates. The loop closes — or repeats.
When agents need a human
Any time an autonomous agent needs approval, input, or a decision before it continues.
- Spec Review — Agent writes a technical spec, stakeholder validates before implementation begins.
- Client Deliverable — Agent generates a report. PM reviews and approves before sending to client.
- Cross-Team Input — Get a decision from another team without sharing your codebase.
- Code Review Without Access — Share changes with a contractor who shouldn't see the repo.
- Architecture Decision — Agent proposes approach with options. Team reviews async, picks an option.
- Webhook-Driven Automation — Agent publishes, waits for webhook notification when feedback arrives, auto-continues.
The missing piece in agent workflows
Other tools let agents create. Profer lets humans review and close the loop.
| Feature | Profer | Claude Artifacts | Google Docs | GitHub Issues |
|---|---|---|---|---|
| Works with any MCP client | ✓ | ✗ | ✗ | ✗ |
| Agent can publish natively | ✓ | ✓ | ✗ | ✗ |
| Shareable URL | ✓ | ✓ | ✓ | ✓ |
| Structured feedback | ✓ | ✗ | ✗ | ✗ |
| Agent reads feedback back | ✓ | ✗ | ✗ | ✗ |
| No login for reviewers | ✓ | ✗ | ✗ | ✓ |
| Version tracking | ✓ | ✓ | ✓ | ✓ |
| Webhook notifications | ✓ | ✗ | ✗ | ✓ |
| No repo access needed | ✓ | ✓ | ✓ | ✗ |
Pricing
Profer is in alpha. Everything is free while we build.
Alpha
Full access while we're in alpha
- ✓Unlimited pages
- ✓All question types
- ✓Public URLs
- ✓MCP tools
- ✓Chat + Canvas
- ✓Realtime agent replies