What is the best nbviewer alternative for a private notebook?

nbviewer needs a publicly reachable URL, so it structurally cannot render a private notebook. The alternatives split by what you actually need: publish converted HTML with identity-based access and comments (Comma), keep it in a hosted notebook platform (Deepnote, Hex, Colab), or render it inside your own infrastructure (nbconvert plus whatever you already host).

Can nbviewer render a notebook from a private GitHub repo?

No. nbviewer fetches the notebook over the public internet — a public repo, a gist, or a public URL. A private repo returns 404 to it. Any workaround that makes it work has made the notebook public, which is usually the thing you were trying to avoid.

Does nbviewer keep interactive plots working?

Plotly, Bokeh and Altair output generally survives because it is self-contained JavaScript in the notebook's stored output. Widgets bound to a live kernel do not, on nbviewer or anywhere else that renders a static document — there is no Python running behind it.

What is the simplest way to share a notebook that people can comment on?

Convert once — `jupyter nbconvert --to html --embed-images` — and publish the file to a review layer. Comma gives it a URL with visibility levels (private, team, email-domain, named reviewers), anchored comment threads on the rendered output, and a revision per rerun on the same link.

nbviewer alternatives

nbviewer is one of the most quietly useful things in the Jupyter ecosystem, and it has one hard constraint that no amount of configuration removes: it fetches your notebook over the public internet. A gist, a public repo, a public URL. That is the deal.

Which means the moment your notebook contains customer numbers, an internal metric, or an unreleased result, nbviewer stops being an option — and the usual fallback is worse. The notebook gets emailed as a .ipynb, opened by nobody, and the conclusions get retyped into a doc where the chart is a screenshot.

What people are actually replacing

Three different needs get called "an nbviewer alternative":

Private rendering. Same experience, but the notebook is not public.
A review surface. nbviewer is read-only by design; there is nowhere to ask "is this cohort excluding trials?"
Freshness. The rendered page is whatever was committed. A rerun means a new commit, a new URL in the message, and a colleague reading last week's numbers.

The shortlist

1. Comma — convert once, publish with access control and comments

Run jupyter nbconvert --to html --embed-images analysis.ipynb, then drag the file in or POST it from the notebook's last cell. What you get that nbviewer cannot give you:

Visibility levels — private, team, email-domain-gated, named reviewers, or public. Not a shared password: identity.
Anchored comment threads on the rendered output, so the question lands on the cohort table rather than in a channel.
A stable URL with revisions — rerun, PATCH the same report id, and the link you sent in January still works and shows what changed.
Publishing from an agent or CI — one POST, or one MCP call, which matters now that a lot of notebooks are produced by something that is not a person.

Not for: running the notebook. There is no kernel; this is the document, after execution. Walkthrough: share a Jupyter notebook as HTML.

Pricing: Free — unlimited reports, commenters, and revision history. Viewers are free and unlimited.

2. GitHub / GitLab's built-in renderer — fine, until it isn't

Both render .ipynb in-repo, including private repos, which genuinely covers the private case for people who already live in the repo. The limits show up fast: large notebooks fall back to raw JSON, interactive output is stripped, review comments land on JSON diff lines instead of on the chart, and every rerun is a commit.

3. Deepnote, Hex, Colab — keep it a live notebook

If what you want is for the reader to run things, a hosted notebook platform is the honest answer, with real collaboration and real access control. You are buying a workspace, not a share link, and pricing follows. See Hex alternatives and the Hex comparison for where that line sits.

4. Self-hosted nbconvert + static hosting

nbconvert in CI, output to S3, Cloudflare Pages or GitHub Pages behind whatever gate you operate. Maximum control, and you own the auth, the invalidation and the review problem — which stays unsolved, because static hosting has no comment layer. See GitHub Pages alternatives.

At a glance

Tool	Private notebooks	Review / comments	Freshness on one URL	Interactive output
Comma	Yes, identity-based	Anchored threads	Revisions on one id	Yes, sandboxed
nbviewer	No — public URL only	No	Re-render per commit	Yes
GitHub renderer	Yes, via repo access	On diff lines	Per commit	Mostly stripped
Deepnote / Hex	Yes	Yes, in-product	Live	Yes, live
Static hosting	You build it	No	Per deploy	Yes

Checked August 2026. Verify current plans before committing.

How to choose

Public, open-source notebook? nbviewer is still excellent. Keep it.
Private analysis that needs a second opinion? Convert to HTML and publish it somewhere with identity-based access and a comment layer — that is the job Comma does.
Reader needs to run the code? A hosted notebook platform, not a viewer.