BrunoThe Git-native API client. Local-first, open source.
briefroom
Built for AI agents. Reviewed by humans.
6 followers
Built for AI agents. Reviewed by humans.
6 followers
Your AI agent can learn briefroom from /llms.txt and use it end-to-end β deploy HTML in one command, share a live URL, and pull human comments back as actionable Markdown. Humans review right on the page: click any element to comment, no account needed, works on mobile. Your agent applies the fixes and redeploys to the same URL. The loop is closed.
Free Options
Launch Team / Built With
Hi Product Hunt!
AI coding agents are great at generating HTML β proposals, mockups, dashboards β but terrible at sharing them. You end up zipping files, screenshotting, or spinning up hosting just to get one client review.
briefroom closes that loop:
Deploy in 30s β drag & drop, or npx @briefroom/cli deploy ./ straight from your agent
Comments on the page β clients tap any element and leave feedback (works on mobile)
Feedback β prompts β pull all comments as LLM-ready Markdown and let your agent apply the fixes
Secure by default β isolated delivery domain, malware scanning, password & invite-only sharing (Pro)
We're launching with a Founders Edition: lifetime Pro for Β₯4,800 (~$30), first 50 people only.
Built in Japan, works in English & Japanese. I'd love your feedback β especially from anyone running AI-driven client work.
Curious how this handles pages with a lot of dynamic JS or auth-gated routes since the share link just has to work in someone else's browser.
@mkoramaz32336Β
Great question! briefroom is intentionally scoped to static output β
the HTML/CSS/JS your agent generates is served as-is from an isolated
sandbox domain (with a strict CSP injected), so self-contained
client-side JS works fine in any viewer's browser.
There's no server runtime though, so SSR or auth-gated app routes are
out of scope by design β for a full app you'd still want a real staging
deploy. briefroom covers the step before that: proposals, mockups, and
LPs that need client eyes fast.
The "only the right people can open it" part is handled at the
share-link level instead: expiring links, password protection (Pro) , and
email-invite-only sharing (Pro) β so the client never needs an account.
Happy to go deeper on the CSP/sandboxing setup if you're curious!
Finally a clean way to get client feedback on AI-built pages without endless screenshots. The markdown export straight into my agent loop is a real time saver, tried it on a quick prototype and it just worked.
@kenan78751Β
Thank you for actually trying it β comments like this make my day! π
The Markdown export is exactly the piece I built briefroom around:
every comment comes back structured (which element, who said it, what
they said), so your agent can apply all the fixes in one pass instead
of you re-explaining each screenshot.
Small tip for your agent loop: point your agent at
https://briefroom.net/llms.txt and it learns the whole CLI workflow
by itself β most people just add one line to their CLAUDE.md / AGENTS.md.
Curious which agent you're running it with (Claude Code / Codex /
Cursor)? And if you hit any friction on that prototype, I'd genuinely
love to hear it β solo dev here, so feedback goes straight into next
week's build.
The handoff between cursor output and client feedback has always been janky for me, so this looks genuinely useful. Love that the comments come back as markdown the agent can actually act on instead of another doc to wrangle.
@c_cank75781Β
Thanks! That janky handoff is exactly why I built this. It works the same with Cursor β deploy via the CLI, then `feedback pull` returns the comments as markdown your agent can act on directly. Would love to hear how it fits your flow if you give it a spin.