Nativ Docs is the collaborative review room for AI-made work. Claude, Codex, and ChatGPT can draft specs, docs, and HTML decks in seconds. But sharing a link only lets people see the work—Nativ lets teams participate in it. • Render Markdown specs & standalone HTML presentations • Highlight exact passages to discuss in the margin • Capture decisions to feed the next prompt or commit • Connect directly to AI workflows via MCP
Nativ solves a very real and increasingly important problem: making AI-generated work easy for teams to review, discuss, and improve together. The feedback loop between human collaboration and AI iteration feels particularly powerful and well thought out.
What needs improvement
Honestly, the core experience feels very polished, and I haven't found any major flaws so far. If anything, deeper integrations with tools like ChatGPT, Claude, Google Drive, and Slack could make an already great workflow even more seamless.
vs Alternatives
I considered using Google Docs, Notion, or collaborating directly through ChatGPT/Claude and Slack. While those tools work well individually, Nativ feels much more purpose-built for the AI-first workflow.
As a product marketer collaboration is critical. Adding comments and easy sharing makes the tool easy to use.
Ratings
Ease of use
Reliability
Value for money
Customization
Report
11 views
Maker
📌
Hey Product Hunt! 👋
I’m thrilled to launch Nativ Docs. The collaborative review room for AI-made work.
The Problem We can now generate complete technical specs, strategy docs, and interactive HTML presentations in seconds with ai agents such as Codex and Claude code.
The bottleneck isn’t creation anymore, it’s reviewing: - Sharing a raw chat link is read-only. - Pasting Markdown into Slack loses formatting and scatters discussion. - Moving everything into heavy doc suites strips out code blocks and interactive layouts.
How Nativ Docs Works 1. Bring in the work: Load rendered Markdown or self-contained HTML artifacts (or push them straight from Claude Code / Codex via our MCP server). 2. Review in context: Invite your team to read the rendered artifact. 3. Anchor comments: Highlight exact text to spin up focused discussion threads in the margin. 4. Resolve to decisions: Capture agreed-upon changes so feedback directly improves the next draft or commit.
Try it out You can try Nativ Docs today with any Markdown spec or HTML artifact. We also built an open MCP server (`@nativdocs/mcp`) so your AI agents can spin up review rooms autonomously.
to answer your question directly - right now my team's "review" step is just pasting the AI's markdown into a Google Doc and leaving suggestion-mode comments, which is exactly the formatting-loss problem you're describing. the part I'm curious about is step 4, resolve to decisions - once comments are marked resolved, does that turn into structured context that gets fed back into Claude Code or Codex for the next generation automatically via the MCP server, or is copying the resolved feedback back into the coding agent still a manual step on our end
@ankur_rustagi1 got it, so the round trip already goes through mcp today rather than a manual copy-paste back into the agent - that's better than I assumed. the memory layer goal makes sense as the natural next step then, since right now each doc's resolved comments are presumably scoped to that one review session. curious if that's the shape you're picturing - the agent querying past resolved decisions across docs before it starts a new pass, so it stops re-litigating things that were already settled in a prior review
Report
Maker
@galdayan yes the review/comment loop is already present. Let me know if you have any feedback to improve it.
Both the approved docs and resolved comments can serve as the org memory. And memory can be used directly to prep/review the doc before any human looks at it.
@ankur_rustagi1 makes sense. one bit of feedback since you asked - the thing I'd actually want as a user is visibility into WHY a comment got marked resolved, not just that it did. if the agent is going to lean on resolved comments as memory later, a one-line "resolved because X" trail would matter a lot more than the resolution itself once you're several docs deep and trying to trust that memory instead of re-reading everything
Hey Product Hunt! 👋
I’m thrilled to launch Nativ Docs. The collaborative review room for AI-made work.
The Problem
We can now generate complete technical specs, strategy docs, and interactive HTML presentations in seconds with ai agents such as Codex and Claude code.
The bottleneck isn’t creation anymore, it’s reviewing:
- Sharing a raw chat link is read-only.
- Pasting Markdown into Slack loses formatting and scatters discussion.
- Moving everything into heavy doc suites strips out code blocks and interactive layouts.
How Nativ Docs Works
1. Bring in the work: Load rendered Markdown or self-contained HTML artifacts (or push them straight from Claude Code / Codex via our MCP server).
2. Review in context: Invite your team to read the rendered artifact.
3. Anchor comments: Highlight exact text to spin up focused discussion threads in the margin.
4. Resolve to decisions: Capture agreed-upon changes so feedback directly improves the next draft or commit.
Try it out
You can try Nativ Docs today with any Markdown spec or HTML artifact. We also built an open MCP server (`@nativdocs/mcp`) so your AI agents can spin up review rooms autonomously.
Sample presentation: https://www.nativdocs.co/s/KzuDX...
Would love to hear how your team currently review AI-generated specs and artifacts?
We'll be here to answer questions and chatting in the comments! 🚀
Dial
to answer your question directly - right now my team's "review" step is just pasting the AI's markdown into a Google Doc and leaving suggestion-mode comments, which is exactly the formatting-loss problem you're describing. the part I'm curious about is step 4, resolve to decisions - once comments are marked resolved, does that turn into structured context that gets fed back into Claude Code or Codex for the next generation automatically via the MCP server, or is copying the resolved feedback back into the coding agent still a manual step on our end
@galdayan thanks for checking us out.
Amazing question/suggestion on the resolved comments. My future goal is to serve all the docs and resolved comments as a memory layer.
Currently, the comments are fetched, docs can be edited and comments can be resolved through mcp. Motivation was the code review flow in git.
Dial
@ankur_rustagi1 got it, so the round trip already goes through mcp today rather than a manual copy-paste back into the agent - that's better than I assumed. the memory layer goal makes sense as the natural next step then, since right now each doc's resolved comments are presumably scoped to that one review session. curious if that's the shape you're picturing - the agent querying past resolved decisions across docs before it starts a new pass, so it stops re-litigating things that were already settled in a prior review
@galdayan yes the review/comment loop is already present. Let me know if you have any feedback to improve it.
Both the approved docs and resolved comments can serve as the org memory. And memory can be used directly to prep/review the doc before any human looks at it.
Dial
@ankur_rustagi1 makes sense. one bit of feedback since you asked - the thing I'd actually want as a user is visibility into WHY a comment got marked resolved, not just that it did. if the agent is going to lean on resolved comments as memory later, a one-line "resolved because X" trail would matter a lot more than the resolution itself once you're several docs deep and trying to trust that memory instead of re-reading everything
Revfx Account Lookalikes
Congrats on the launch ! I have added this as a skill which is very helpful when I am trying to navigate a new code base
@saurabh_dhupar thanks! Codebase learning through visuals is much easier. The MD files rendering in Nativ supporta mermaid diagrams.